├── README.md
├── zind-airqo-final-blend.ipynb
├── darius-model-1.ipynb
├── darius-model-2.ipynb
└── zindi-airqo-cnn-quick.ipynb


/README.md:
--------------------------------------------------------------------------------
1 | # zindi-airqo
2 | 
3 | 
4 | 
5 | 1. Run the 4 notebooks 'darius-model-1.ipynb', 'darius-model-2.ipynb',  'model-3.ipynb' and
6 | 'zindi-airqo-cnn-quick.ipynb' in any order. The outputs generated by the notebooks are kept already placed in the current folder.
7 | 2. Run the final notebook 'zind-airqo-final-blend.ipynb' only after running the above 4 notebooks in any order to get the final output 'zindi_airqo_final_sub.csv'.
8 | 


--------------------------------------------------------------------------------
/zind-airqo-final-blend.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 1,
  6 |    "metadata": {
  7 |     "collapsed": true
  8 |    },
  9 |    "outputs": [],
 10 |    "source": [
 11 |     "import numpy as np\n",
 12 |     "import pandas as pd"
 13 |    ]
 14 |   },
 15 |   {
 16 |    "cell_type": "code",
 17 |    "execution_count": 2,
 18 |    "metadata": {},
 19 |    "outputs": [
 20 |     {
 21 |      "name": "stdout",
 22 |      "output_type": "stream",
 23 |      "text": [
 24 |       "airqo-ugandan-air-quality-forecast-challenge #2 3 STRoNG !!!.zip\n",
 25 |       "cnn_preds.csv\n",
 26 |       "darius-model-1.ipynb\n",
 27 |       "darius-model-2.ipynb\n",
 28 |       "input\n",
 29 |       "model-3.ipynb\n",
 30 |       "model_12_blend.csv\n",
 31 |       "model_2.csv\n",
 32 |       "model_3.csv\n",
 33 |       "readme.txt\n",
 34 |       "zind-airqo-final-blend.ipynb\n",
 35 |       "zindi-airqo-cnn-quick.ipynb\n",
 36 |       "zindi_airqo_final_sub.csv\n"
 37 |      ]
 38 |     }
 39 |    ],
 40 |    "source": [
 41 |     "!ls"
 42 |    ]
 43 |   },
 44 |   {
 45 |    "cell_type": "code",
 46 |    "execution_count": 3,
 47 |    "metadata": {
 48 |     "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0",
 49 |     "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a",
 50 |     "collapsed": true
 51 |    },
 52 |    "outputs": [],
 53 |    "source": [
 54 |     "df = pd.read_csv('model_2.csv')[['ID']]\n",
 55 |     "df = pd.merge(df, pd.read_csv('model_12_blend.csv').rename({'target': 'A'}, axis=1), on = 'ID', how='left')\n",
 56 |     "df = pd.merge(df, pd.read_csv('model_2.csv').rename({'target': 'B'}, axis=1), on = 'ID', how='left')\n",
 57 |     "df = pd.merge(df, pd.read_csv('model_3.csv').rename({'target': 'C'}, axis=1), on = 'ID', how='left')\n",
 58 |     "df = pd.merge(df, pd.read_csv('cnn_preds.csv').rename({'target': 'D'}, axis=1), on = 'ID', how='left')"
 59 |    ]
 60 |   },
 61 |   {
 62 |    "cell_type": "code",
 63 |    "execution_count": 4,
 64 |    "metadata": {},
 65 |    "outputs": [
 66 |     {
 67 |      "data": {
 68 |       "text/html": [
 69 |        "<div>\n",
 70 |        "<style scoped>\n",
 71 |        "    .dataframe tbody tr th:only-of-type {\n",
 72 |        "        vertical-align: middle;\n",
 73 |        "    }\n",
 74 |        "\n",
 75 |        "    .dataframe tbody tr th {\n",
 76 |        "        vertical-align: top;\n",
 77 |        "    }\n",
 78 |        "\n",
 79 |        "    .dataframe thead th {\n",
 80 |        "        text-align: right;\n",
 81 |        "    }\n",
 82 |        "</style>\n",
 83 |        "<table border=\"1\" class=\"dataframe\">\n",
 84 |        "  <thead>\n",
 85 |        "    <tr style=\"text-align: right;\">\n",
 86 |        "      <th></th>\n",
 87 |        "      <th>A</th>\n",
 88 |        "      <th>B</th>\n",
 89 |        "      <th>C</th>\n",
 90 |        "      <th>D</th>\n",
 91 |        "    </tr>\n",
 92 |        "  </thead>\n",
 93 |        "  <tbody>\n",
 94 |        "    <tr>\n",
 95 |        "      <th>A</th>\n",
 96 |        "      <td>1.000000</td>\n",
 97 |        "      <td>0.972463</td>\n",
 98 |        "      <td>0.981905</td>\n",
 99 |        "      <td>0.937631</td>\n",
100 |        "    </tr>\n",
101 |        "    <tr>\n",
102 |        "      <th>B</th>\n",
103 |        "      <td>0.972463</td>\n",
104 |        "      <td>1.000000</td>\n",
105 |        "      <td>0.966836</td>\n",
106 |        "      <td>0.932886</td>\n",
107 |        "    </tr>\n",
108 |        "    <tr>\n",
109 |        "      <th>C</th>\n",
110 |        "      <td>0.981905</td>\n",
111 |        "      <td>0.966836</td>\n",
112 |        "      <td>1.000000</td>\n",
113 |        "      <td>0.941585</td>\n",
114 |        "    </tr>\n",
115 |        "    <tr>\n",
116 |        "      <th>D</th>\n",
117 |        "      <td>0.937631</td>\n",
118 |        "      <td>0.932886</td>\n",
119 |        "      <td>0.941585</td>\n",
120 |        "      <td>1.000000</td>\n",
121 |        "    </tr>\n",
122 |        "  </tbody>\n",
123 |        "</table>\n",
124 |        "</div>"
125 |       ],
126 |       "text/plain": [
127 |        "          A         B         C         D\n",
128 |        "A  1.000000  0.972463  0.981905  0.937631\n",
129 |        "B  0.972463  1.000000  0.966836  0.932886\n",
130 |        "C  0.981905  0.966836  1.000000  0.941585\n",
131 |        "D  0.937631  0.932886  0.941585  1.000000"
132 |       ]
133 |      },
134 |      "execution_count": 4,
135 |      "metadata": {},
136 |      "output_type": "execute_result"
137 |     }
138 |    ],
139 |    "source": [
140 |     "df[['A', 'B', 'C', 'D']].corr()"
141 |    ]
142 |   },
143 |   {
144 |    "cell_type": "code",
145 |    "execution_count": 5,
146 |    "metadata": {},
147 |    "outputs": [
148 |     {
149 |      "name": "stderr",
150 |      "output_type": "stream",
151 |      "text": [
152 |       "C:\\Anaconda5\\lib\\site-packages\\ipykernel_launcher.py:2: SettingWithCopyWarning: \n",
153 |       "A value is trying to be set on a copy of a slice from a DataFrame.\n",
154 |       "Try using .loc[row_indexer,col_indexer] = value instead\n",
155 |       "\n",
156 |       "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
157 |       "  \n"
158 |      ]
159 |     }
160 |    ],
161 |    "source": [
162 |     "final_sub_df = df[['ID']]\n",
163 |     "final_sub_df['target'] = ((df['A']*0.6 + df['B']*0.4)*0.35 + df['C']*0.65)*0.85 + df['D']*0.15\n",
164 |     "final_sub_df.to_csv('zindi_airqo_final_sub.csv', index=False)"
165 |    ]
166 |   },
167 |   {
168 |    "cell_type": "code",
169 |    "execution_count": 6,
170 |    "metadata": {},
171 |    "outputs": [
172 |     {
173 |      "name": "stderr",
174 |      "output_type": "stream",
175 |      "text": [
176 |       "C:\\Anaconda5\\lib\\site-packages\\ipykernel_launcher.py:2: SettingWithCopyWarning: \n",
177 |       "A value is trying to be set on a copy of a slice from a DataFrame.\n",
178 |       "Try using .loc[row_indexer,col_indexer] = value instead\n",
179 |       "\n",
180 |       "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
181 |       "  \n"
182 |      ]
183 |     }
184 |    ],
185 |    "source": [
186 |     "final_sub_df = df[['ID']]\n",
187 |     "final_sub_df['target'] = (df['A']*0.6 + df['B']*0.4)\n",
188 |     "final_sub_df.to_csv('zindi_airqo_model_1_and_2_only.csv', index=False)"
189 |    ]
190 |   },
191 |   {
192 |    "cell_type": "code",
193 |    "execution_count": null,
194 |    "metadata": {
195 |     "collapsed": true
196 |    },
197 |    "outputs": [],
198 |    "source": []
199 |   }
200 |  ],
201 |  "metadata": {
202 |   "kernelspec": {
203 |    "display_name": "Python 3",
204 |    "language": "python",
205 |    "name": "python3"
206 |   },
207 |   "language_info": {
208 |    "codemirror_mode": {
209 |     "name": "ipython",
210 |     "version": 3
211 |    },
212 |    "file_extension": ".py",
213 |    "mimetype": "text/x-python",
214 |    "name": "python",
215 |    "nbconvert_exporter": "python",
216 |    "pygments_lexer": "ipython3",
217 |    "version": "3.6.3"
218 |   }
219 |  },
220 |  "nbformat": 4,
221 |  "nbformat_minor": 4
222 | }
223 | 


--------------------------------------------------------------------------------
/darius-model-1.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": null,
  6 |    "metadata": {
  7 |     "collapsed": true,
  8 |     "id": "pk5WiopT9oJw",
  9 |     "outputId": "446b6db5-1cbe-4beb-e1e2-4bcdd512bab6"
 10 |    },
 11 |    "outputs": [],
 12 |    "source": [
 13 |     "# installing catboost\n",
 14 |     "# Catboost == 0.22 was the version of catboost at the start of this competition\n",
 15 |     "!pip install catboost==0.22 --quiet"
 16 |    ]
 17 |   },
 18 |   {
 19 |    "cell_type": "code",
 20 |    "execution_count": null,
 21 |    "metadata": {
 22 |     "collapsed": true,
 23 |     "id": "jVXG4QlTuVYr"
 24 |    },
 25 |    "outputs": [],
 26 |    "source": [
 27 |     "# Importing libraries\n",
 28 |     "import pandas as pd\n",
 29 |     "import numpy as np\n",
 30 |     "import warnings\n",
 31 |     "import joblib\n",
 32 |     "\n",
 33 |     "from tqdm import tqdm, tqdm_notebook\n",
 34 |     "from functools import reduce\n",
 35 |     "from time import time\n",
 36 |     "\n",
 37 |     "from catboost import CatBoostRegressor, CatBoostClassifier\n",
 38 |     "from sklearn.utils import shuffle\n",
 39 |     "\n",
 40 |     "pd.set_option('display.max_rows', 1000)           \n",
 41 |     "warnings.filterwarnings('ignore')"
 42 |    ]
 43 |   },
 44 |   {
 45 |    "cell_type": "code",
 46 |    "execution_count": null,
 47 |    "metadata": {
 48 |     "collapsed": true,
 49 |     "id": "VJjtuM3kvCAT"
 50 |    },
 51 |    "outputs": [],
 52 |    "source": [
 53 |     "# Loading data\n",
 54 |     "train = pd.read_csv('./input/Train.csv')\n",
 55 |     "test = pd.read_csv('./input/Test.csv')"
 56 |    ]
 57 |   },
 58 |   {
 59 |    "cell_type": "code",
 60 |    "execution_count": null,
 61 |    "metadata": {
 62 |     "collapsed": true
 63 |    },
 64 |    "outputs": [],
 65 |    "source": [
 66 |     "start = time()"
 67 |    ]
 68 |   },
 69 |   {
 70 |    "cell_type": "code",
 71 |    "execution_count": null,
 72 |    "metadata": {
 73 |     "collapsed": true,
 74 |     "id": "4MhDKOpYF9dm"
 75 |    },
 76 |    "outputs": [],
 77 |    "source": [
 78 |     "# Separating the target variable from the training dataframe\n",
 79 |     "#\n",
 80 |     "target = train.target\n",
 81 |     "\n",
 82 |     "# Aligning the train and test dataframes\n",
 83 |     "#\n",
 84 |     "train, test = train.align(test, join='inner', axis=1)\n",
 85 |     "\n",
 86 |     "# Creating a separator column to both train and test,\n",
 87 |     "# This is to be used in separation\n",
 88 |     "#\n",
 89 |     "train['separator'] = 0\n",
 90 |     "test['separator'] = 1\n",
 91 |     "\n",
 92 |     "# Combing the train and test dataframes together\n",
 93 |     "#\n",
 94 |     "comb = pd.concat([train, test])\n",
 95 |     "\n",
 96 |     "# Creating a function to replace all spaces in the dataframe with np.nan\n",
 97 |     "#\n",
 98 |     "def replace_nan(x):\n",
 99 |     "    if x == \" \":\n",
100 |     "        return np.nan\n",
101 |     "    else:\n",
102 |     "        return float(x)\n",
103 |     "\n",
104 |     "# Creating a list of the main columns\n",
105 |     "#\n",
106 |     "main_cols = [\"temp\", \"precip\", \"rel_humidity\", \"wind_dir\", \"wind_spd\", \"atmos_press\"]\n",
107 |     "\n",
108 |     "# Replacing spaces with np.nan\n",
109 |     "#\n",
110 |     "for col in main_cols: \n",
111 |     "    comb[col] = comb[col].apply(lambda x: [replace_nan(X) for X in x.replace(\"nan\", \" \").split(\",\")])\n",
112 |     "\n",
113 |     "def make_columns(feature):\n",
114 |     "    return [f\"{feature}_{i}\" for i in range(1, 122)]\n",
115 |     "    \n",
116 |     "# Generating dataframes of hours for each main column\n",
117 |     "#\n",
118 |     "comb_temp         = pd.DataFrame([x for x in comb.temp],         columns=make_columns('temp'))\n",
119 |     "comb_precip       = pd.DataFrame([x for x in comb.precip],       columns=make_columns('precip'))\n",
120 |     "comb_rel_humidity = pd.DataFrame([x for x in comb.rel_humidity], columns=make_columns('rel_humidity'))\n",
121 |     "comb_wind_dir     = pd.DataFrame([x for x in comb.wind_dir],     columns=make_columns('wind_dir'))\n",
122 |     "comb_wind_spd     = pd.DataFrame([x for x in comb.wind_spd],     columns=make_columns('wind_spd'))\n",
123 |     "comb_atmos_press  = pd.DataFrame([x for x in comb.atmos_press],  columns=make_columns('atmos_press'))\n",
124 |     "\n",
125 |     "comb_temp['ID'], comb_precip['ID'], comb_rel_humidity['ID'], comb_wind_dir['ID'], comb_wind_spd['ID'], comb_atmos_press['ID'] = [list(comb.ID)] * 6\n",
126 |     "\n",
127 |     "# Combining the generated dataframes together\n",
128 |     "#\n",
129 |     "comb_dfs = [comb, comb_temp, comb_precip, comb_rel_humidity, comb_wind_dir, comb_wind_spd, comb_atmos_press]\n",
130 |     "comb = reduce(lambda  left, right: pd.merge(left, right, on=['ID'], how='outer'), comb_dfs)\n",
131 |     "comb.drop(main_cols, axis=1, inplace=True)\n",
132 |     "df = comb.copy()"
133 |    ]
134 |   },
135 |   {
136 |    "cell_type": "code",
137 |    "execution_count": null,
138 |    "metadata": {
139 |     "code_folding": [],
140 |     "collapsed": true,
141 |     "id": "3FLjxrAXYbRV",
142 |     "outputId": "5d01c7d2-cf39-49ce-d234-6c735a3971d7"
143 |    },
144 |    "outputs": [],
145 |    "source": [
146 |     "# Creating original series for each feature\n",
147 |     "orig_cols_dict = {}\n",
148 |     "weather_cols = ['temp', 'precip', 'rel_humidity', 'wind_dir','wind_spd', 'atmos_press']\n",
149 |     "\n",
150 |     "for w in tqdm_notebook(weather_cols):\n",
151 |     "    selected_cols = [c for c in df.columns if w in c]\n",
152 |     "    orig_cols_dict[w] = pd.Series(selected_cols)"
153 |    ]
154 |   },
155 |   {
156 |    "cell_type": "code",
157 |    "execution_count": null,
158 |    "metadata": {
159 |     "collapsed": true,
160 |     "id": "oAGDsJ92YS2Z",
161 |     "outputId": "3c4077e0-d77c-4feb-e744-4ef904531c04"
162 |    },
163 |    "outputs": [],
164 |    "source": [
165 |     "# Aggregating features per hour\n",
166 |     "\n",
167 |     "for w in tqdm_notebook(weather_cols):\n",
168 |     "    tmp_df = pd.DataFrame()\n",
169 |     "    tmp_df['weather_col_orig'] = orig_cols_dict[w]\n",
170 |     "    tmp_df['hours_since_start'] = tmp_df['weather_col_orig'].apply(lambda x: x.split('_')[-1]).astype('int')\n",
171 |     "    tmp_df['hour_of_day'] = tmp_df['hours_since_start'] % 24\n",
172 |     "\n",
173 |     "    for hour in range(1, 25):\n",
174 |     "        selected_cols = tmp_df[tmp_df['hour_of_day'] == hour]['weather_col_orig'].tolist()\n",
175 |     "        df_cols = df[selected_cols] # factorizing this part\n",
176 |     "        \n",
177 |     "        df[f'{w}_hour_{hour}_mean'] = df_cols.mean(axis=1)\n",
178 |     "        df[f'{w}_hour_{hour}_min'] = df_cols.min(axis=1)\n",
179 |     "        df[f'{w}_hour_{hour}_max'] = df_cols.max(axis=1)\n",
180 |     "        df[f'{w}_hour_{hour}_range'] = df[f'{w}_hour_{hour}_max'] - df[f'{w}_hour_{hour}_min']\n",
181 |     "        df[f'{w}_hour_{hour}_skew'] = df_cols.skew()\n",
182 |     "        df[f'{w}_hour_{hour}_kurt'] = df_cols.kurt()\n",
183 |     "\n",
184 |     "        if hour - 3 > 0 and hour % 3 == 0:\n",
185 |     "            df[f'{w}_hour_{hour}_prev_hour_mean_diff'] = df[f'{w}_hour_{hour}_mean'] - df[f'{w}_hour_{hour - 3}_mean']\n",
186 |     "        if hour - 5 > 0 and hour % 3 == 0:\n",
187 |     "            df[f'{w}_hour_{hour}_prev_hour_mean_diff_5'] = df[f'{w}_hour_{hour}_mean'] - df[f'{w}_hour_{hour - 5}_mean']\n"
188 |    ]
189 |   },
190 |   {
191 |    "cell_type": "code",
192 |    "execution_count": null,
193 |    "metadata": {
194 |     "collapsed": true,
195 |     "id": "pEyFVf1lqInY"
196 |    },
197 |    "outputs": [],
198 |    "source": [
199 |     "comb = df.copy()"
200 |    ]
201 |   },
202 |   {
203 |    "cell_type": "code",
204 |    "execution_count": null,
205 |    "metadata": {
206 |     "collapsed": true,
207 |     "id": "WCp_Ukh-NTso",
208 |     "outputId": "104e385d-04c6-47af-8bec-582bc0fdfbac"
209 |    },
210 |    "outputs": [],
211 |    "source": [
212 |     "comb.head()"
213 |    ]
214 |   },
215 |   {
216 |    "cell_type": "code",
217 |    "execution_count": null,
218 |    "metadata": {
219 |     "collapsed": true,
220 |     "id": "xzn_w6FYftB8",
221 |     "outputId": "33af157e-0c05-4be9-fd8f-3290f64fcd47"
222 |    },
223 |    "outputs": [],
224 |    "source": [
225 |     "# Creating aggregation features for each variable\n",
226 |     "aggs = ['mean', 'std', 'var', 'kurt', 'skew', 'max', 'median', 'sum', 'mode', 'sem', 'min']\n",
227 |     "\n",
228 |     "for col in tqdm_notebook(main_cols):\n",
229 |     "    for ag in tqdm(aggs):\n",
230 |     "        selected_cols = [x for x in comb.columns if x.startswith(col)]\n",
231 |     "\n",
232 |     "        if ag == 'mode':\n",
233 |     "            aggregate = comb[selected_cols].agg(ag, axis=1)[0]\n",
234 |     "        else:\n",
235 |     "            aggregate = comb[selected_cols].agg(ag, axis=1)\n",
236 |     "        \n",
237 |     "        comb[col[0] + col[-1] + '_' + ag] = aggregate\n",
238 |     "\n",
239 |     "# Creating separate dataframes for each variable\n",
240 |     "# Creating a list of columns for each separate dataframe\n",
241 |     "temp_cols = [x for x in comb.columns if x.startswith('temp')]\n",
242 |     "temp = comb[temp_cols]\n",
243 |     "\n",
244 |     "precip_cols = [x for x in comb.columns if x.startswith('precip')]\n",
245 |     "precip = comb[precip_cols]\n",
246 |     "\n",
247 |     "humid_cols = [x for x in comb.columns if x.startswith('rel_humidity')]\n",
248 |     "humid = comb[humid_cols]\n",
249 |     "\n",
250 |     "wind_dir_cols = [x for x in comb.columns if x.startswith('wind_dir')]\n",
251 |     "wind_dir = comb[wind_dir_cols]\n",
252 |     "\n",
253 |     "wind_spd_cols = [x for x in comb.columns if x.startswith('wind_spd')]\n",
254 |     "wind_spd  = comb[wind_spd_cols]\n",
255 |     "\n",
256 |     "atmp_cols = [x for x in comb.columns if x.startswith('atmos_press')]\n",
257 |     "atmp = comb[atmp_cols]\n",
258 |     "\n",
259 |     "fill_cols = comb.columns"
260 |    ]
261 |   },
262 |   {
263 |    "cell_type": "code",
264 |    "execution_count": null,
265 |    "metadata": {
266 |     "collapsed": true,
267 |     "id": "qdVrf3ZmcFZn"
268 |    },
269 |    "outputs": [],
270 |    "source": [
271 |     "# Generating new features, by adding each variable per hour\n",
272 |     "for x, y, z, a, b in zip(temp.columns, precip.columns, humid.columns, wind_spd.columns, atmp.columns):\n",
273 |     "    comb['add_tp' + y[-4:]] = temp[x] + precip[y] + humid[z] + wind_spd[a] + atmp[b]"
274 |    ]
275 |   },
276 |   {
277 |    "cell_type": "code",
278 |    "execution_count": null,
279 |    "metadata": {
280 |     "collapsed": true,
281 |     "id": "IU2Z7uDigrgW"
282 |    },
283 |    "outputs": [],
284 |    "source": [
285 |     "# Filling missing values using forward fill\n",
286 |     "comb = comb.ffill(axis=1)"
287 |    ]
288 |   },
289 |   {
290 |    "cell_type": "code",
291 |    "execution_count": null,
292 |    "metadata": {
293 |     "collapsed": true,
294 |     "id": "aQTxY80kZ6rH",
295 |     "outputId": "a2e1dfe2-2032-4666-9921-e523a4e566e6"
296 |    },
297 |    "outputs": [],
298 |    "source": [
299 |     "comb.head()"
300 |    ]
301 |   },
302 |   {
303 |    "cell_type": "code",
304 |    "execution_count": null,
305 |    "metadata": {
306 |     "collapsed": true
307 |    },
308 |    "outputs": [],
309 |    "source": [
310 |     "def apply_qcut(feat):\n",
311 |     "    return pd.qcut(comb[feat], 24, labels=False, duplicates='drop')"
312 |    ]
313 |   },
314 |   {
315 |    "cell_type": "code",
316 |    "execution_count": null,
317 |    "metadata": {
318 |     "collapsed": true
319 |    },
320 |    "outputs": [],
321 |    "source": [
322 |     "other_features = [x for x in comb.columns if x not in ['separator', 'ID', 'location']]\n",
323 |     "\n",
324 |     "# Multiprocessing trick: 15 seconds instead of 7 minutes !\n",
325 |     "binned_data = joblib.Parallel(n_jobs=-1, backend='multiprocessing')(\n",
326 |     "    joblib.delayed(apply_qcut)(feat) for feat in tqdm_notebook(other_features))\n",
327 |     "\n",
328 |     "comb_binned_data = pd.concat(binned_data, axis=1)\n",
329 |     "comb = pd.concat([comb[['separator', 'ID', 'location']], comb_binned_data], axis=1)"
330 |    ]
331 |   },
332 |   {
333 |    "cell_type": "code",
334 |    "execution_count": null,
335 |    "metadata": {
336 |     "collapsed": true
337 |    },
338 |    "outputs": [],
339 |    "source": [
340 |     "comb.head()"
341 |    ]
342 |   },
343 |   {
344 |    "cell_type": "code",
345 |    "execution_count": null,
346 |    "metadata": {
347 |     "collapsed": true,
348 |     "id": "w-60osCYdgjT",
349 |     "outputId": "7dc3c3bd-ef95-4d34-f090-3c7efb05081a"
350 |    },
351 |    "outputs": [],
352 |    "source": [
353 |     "# Separating train and test from the combined dataframe\n",
354 |     "train = comb[comb.separator == 0]\n",
355 |     "test = comb[comb.separator == 1]\n",
356 |     "train.drop('separator', axis=1, inplace=True)\n",
357 |     "test.drop('separator', axis=1, inplace=True)\n",
358 |     "\n",
359 |     "# Creating a list of test ids in the order that they will be trained\n",
360 |     "testA =  test[test.location == 'A']\n",
361 |     "testB =  test[test.location == 'B']\n",
362 |     "testC =  test[test.location == 'C']\n",
363 |     "testD =  test[test.location == 'D']\n",
364 |     "testE =  test[test.location == 'E']\n",
365 |     "\n",
366 |     "tA, tD, tE, tBC = testA.ID, testD.ID, testE.ID, test[(test.location == 'B') | (test.location == 'C')].ID\n",
367 |     "test_id = pd.concat([tA, tD, tE, tBC])\n",
368 |     "\n",
369 |     "# Adding back target to the train set\n",
370 |     "train['target'] = target"
371 |    ]
372 |   },
373 |   {
374 |    "cell_type": "code",
375 |    "execution_count": null,
376 |    "metadata": {
377 |     "collapsed": true
378 |    },
379 |    "outputs": [],
380 |    "source": [
381 |     "end = time()\n",
382 |     "print(f\"Total preprocessing time = {end - start:.1f}\")"
383 |    ]
384 |   },
385 |   {
386 |    "cell_type": "code",
387 |    "execution_count": null,
388 |    "metadata": {
389 |     "collapsed": true,
390 |     "id": "130zceoDarsp"
391 |    },
392 |    "outputs": [],
393 |    "source": [
394 |     "%%time\n",
395 |     "# Creating X and y values\n",
396 |     "X = train.drop(['ID', 'location', 'target'], axis=1)\n",
397 |     "y = target.values\n",
398 |     "\n",
399 |     "# Shuffling the X, y values\n",
400 |     "X, y = shuffle(X, y, random_state=0)\n",
401 |     "tes = test.drop(['ID', 'location'], axis=1)\n",
402 |     "\n",
403 |     "# Traing the model across multiple seeds\n",
404 |     "predictions = []\n",
405 |     "for i in tqdm_notebook(range(25)):\n",
406 |     "    cat = CatBoostRegressor(verbose=False, random_seed=i)\n",
407 |     "    cat.fit(X, y)\n",
408 |     "    \n",
409 |     "    preds = cat.predict(tes)\n",
410 |     "    predictions.append(preds)\n",
411 |     "\n",
412 |     "# Averaging the predictions\n",
413 |     "avg_preds = np.mean(predictions, axis=0)\n",
414 |     "\n",
415 |     "# Post processing of the predictions\n",
416 |     "# This post processing was done with the help of a validation set.\n",
417 |     "# The validation set was adversarial, i.e. we chose the examples from the training set closest to the test set, and applied post processing to it.\n",
418 |     "post_proc = [((((((((((x-0.85)*1.015)-0.85)*1.012)-0.75)*1.0095)-0.55)*1.0065)-0.8)*1.007) for x in avg_preds]\n",
419 |     "post_proc = predzz = [((x-0.85)*1.015) for x in post_proc]\n",
420 |     "\n",
421 |     "# Creating a submission file\n",
422 |     "sub_df = pd.DataFrame({'ID': test.ID, 'target': post_proc})\n",
423 |     "sub_df.to_csv('model_1_1.csv', index=False)"
424 |    ]
425 |   },
426 |   {
427 |    "cell_type": "code",
428 |    "execution_count": null,
429 |    "metadata": {
430 |     "collapsed": true,
431 |     "id": "nrK5eVysjIna",
432 |     "outputId": "8846d85a-7616-49b3-b0db-d87fc502c993"
433 |    },
434 |    "outputs": [],
435 |    "source": [
436 |     "%%time\n",
437 |     "# Creating a list to hold predictions per seed\n",
438 |     "predzz = []\n",
439 |     "for i in tqdm_notebook(range(25), leave=False):\n",
440 |     "    # Creating a list to hold predictions per location\n",
441 |     "    # Training model per location per seed\n",
442 |     "    predictions = []\n",
443 |     "    for area in tqdm_notebook(['A', 'D', 'E'], leave=False):\n",
444 |     "        # Separating training data per location\n",
445 |     "        X = train[train.location == area]\n",
446 |     "        y = X.target\n",
447 |     "        X = X.drop(['ID', 'location', 'target'], axis=1)\n",
448 |     "\n",
449 |     "        # Shuffling data\n",
450 |     "        X, y = shuffle(X, y, random_state=0)\n",
451 |     "\n",
452 |     "        # Separating testing data per location\n",
453 |     "        tes = test[test.location == area]\n",
454 |     "        tes = tes.drop(['ID', 'location'], axis=1)\n",
455 |     "\n",
456 |     "        # Training the model and making predictions per seed, per location\n",
457 |     "        preds = CatBoostRegressor(verbose=False, random_seed=i).fit(X, y).predict(tes)\n",
458 |     "        predictions.extend(preds)\n",
459 |     "\n",
460 |     "    X = train[(train.location == 'B') | (train.location == 'C')]\n",
461 |     "    y = X.target\n",
462 |     "    X = X.drop(['ID', 'location', 'target'], axis=1)\n",
463 |     "    X, y = shuffle(X, y, random_state=0)\n",
464 |     "\n",
465 |     "    tes = test[(test.location == 'B') | (test.location == 'C')]\n",
466 |     "    tes = tes.drop(['ID', 'location'], axis=1)\n",
467 |     "    preds = CatBoostRegressor(verbose=False, random_seed=i).fit(X, y).predict(tes)\n",
468 |     "    predictions.extend(preds)\n",
469 |     "\n",
470 |     "    predzz.append(predictions)\n",
471 |     "\n",
472 |     "# Averaging the predictions\n",
473 |     "preds_av = np.mean(predzz, axis=0)\n",
474 |     "\n",
475 |     "# Post processing of the predictions\n",
476 |     "# This post processing was done with the help of a validation set.\n",
477 |     "# The validation set was adversarial, i.e. we chose the examples from the training set closest to the test set, and applied post processing to it.\n",
478 |     "predz = [((((((((((x-0.85)*1.015)-0.85)*1.012)-0.75)*1.0095)-0.55)*1.0065)-0.8)*1.007) for x in preds_av]\n",
479 |     "predzz = [((x-0.85)*1.015) for x in predz]\n",
480 |     "\n",
481 |     "# Creating a submission file\n",
482 |     "sub_df = pd.DataFrame({'ID': test_id, 'target': predzz})\n",
483 |     "sub_df.to_csv('model_1_2.csv', index = False)"
484 |    ]
485 |   },
486 |   {
487 |    "cell_type": "code",
488 |    "execution_count": 48,
489 |    "metadata": {},
490 |    "outputs": [
491 |     {
492 |      "data": {
493 |       "text/html": [
494 |        "<div>\n",
495 |        "<style scoped>\n",
496 |        "    .dataframe tbody tr th:only-of-type {\n",
497 |        "        vertical-align: middle;\n",
498 |        "    }\n",
499 |        "\n",
500 |        "    .dataframe tbody tr th {\n",
501 |        "        vertical-align: top;\n",
502 |        "    }\n",
503 |        "\n",
504 |        "    .dataframe thead th {\n",
505 |        "        text-align: right;\n",
506 |        "    }\n",
507 |        "</style>\n",
508 |        "<table border=\"1\" class=\"dataframe\">\n",
509 |        "  <thead>\n",
510 |        "    <tr style=\"text-align: right;\">\n",
511 |        "      <th></th>\n",
512 |        "      <th>A</th>\n",
513 |        "      <th>B</th>\n",
514 |        "    </tr>\n",
515 |        "  </thead>\n",
516 |        "  <tbody>\n",
517 |        "    <tr>\n",
518 |        "      <th>A</th>\n",
519 |        "      <td>1.000000</td>\n",
520 |        "      <td>0.979202</td>\n",
521 |        "    </tr>\n",
522 |        "    <tr>\n",
523 |        "      <th>B</th>\n",
524 |        "      <td>0.979202</td>\n",
525 |        "      <td>1.000000</td>\n",
526 |        "    </tr>\n",
527 |        "  </tbody>\n",
528 |        "</table>\n",
529 |        "</div>"
530 |       ],
531 |       "text/plain": [
532 |        "          A         B\n",
533 |        "A  1.000000  0.979202\n",
534 |        "B  0.979202  1.000000"
535 |       ]
536 |      },
537 |      "execution_count": 48,
538 |      "metadata": {},
539 |      "output_type": "execute_result"
540 |     }
541 |    ],
542 |    "source": [
543 |     "blend_df = pd.read_csv('model_1_1.csv')[['ID']]\n",
544 |     "blend_df['A'] = pd.read_csv('model_1_1.csv')['target']\n",
545 |     "blend_df = pd.merge(blend_df, pd.read_csv('model_1_2.csv').rename({'target': 'B'}, axis=1), on = 'ID', how = 'left')\n",
546 |     "blend_df.corr()"
547 |    ]
548 |   },
549 |   {
550 |    "cell_type": "code",
551 |    "execution_count": 53,
552 |    "metadata": {
553 |     "collapsed": true
554 |    },
555 |    "outputs": [],
556 |    "source": [
557 |     "blend_df['target'] = blend_df['A']*0.5 + blend_df['B']*0.5\n",
558 |     "blend_df[['ID', 'target']].to_csv('model_12_blend.csv', index=False)"
559 |    ]
560 |   },
561 |   {
562 |    "cell_type": "code",
563 |    "execution_count": 54,
564 |    "metadata": {},
565 |    "outputs": [
566 |     {
567 |      "data": {
568 |       "text/html": [
569 |        "<div>\n",
570 |        "<style scoped>\n",
571 |        "    .dataframe tbody tr th:only-of-type {\n",
572 |        "        vertical-align: middle;\n",
573 |        "    }\n",
574 |        "\n",
575 |        "    .dataframe tbody tr th {\n",
576 |        "        vertical-align: top;\n",
577 |        "    }\n",
578 |        "\n",
579 |        "    .dataframe thead th {\n",
580 |        "        text-align: right;\n",
581 |        "    }\n",
582 |        "</style>\n",
583 |        "<table border=\"1\" class=\"dataframe\">\n",
584 |        "  <thead>\n",
585 |        "    <tr style=\"text-align: right;\">\n",
586 |        "      <th></th>\n",
587 |        "      <th>A</th>\n",
588 |        "      <th>B</th>\n",
589 |        "      <th>target</th>\n",
590 |        "    </tr>\n",
591 |        "  </thead>\n",
592 |        "  <tbody>\n",
593 |        "    <tr>\n",
594 |        "      <th>A</th>\n",
595 |        "      <td>1.000000</td>\n",
596 |        "      <td>0.979202</td>\n",
597 |        "      <td>0.994761</td>\n",
598 |        "    </tr>\n",
599 |        "    <tr>\n",
600 |        "      <th>B</th>\n",
601 |        "      <td>0.979202</td>\n",
602 |        "      <td>1.000000</td>\n",
603 |        "      <td>0.994813</td>\n",
604 |        "    </tr>\n",
605 |        "    <tr>\n",
606 |        "      <th>target</th>\n",
607 |        "      <td>0.994761</td>\n",
608 |        "      <td>0.994813</td>\n",
609 |        "      <td>1.000000</td>\n",
610 |        "    </tr>\n",
611 |        "  </tbody>\n",
612 |        "</table>\n",
613 |        "</div>"
614 |       ],
615 |       "text/plain": [
616 |        "               A         B    target\n",
617 |        "A       1.000000  0.979202  0.994761\n",
618 |        "B       0.979202  1.000000  0.994813\n",
619 |        "target  0.994761  0.994813  1.000000"
620 |       ]
621 |      },
622 |      "execution_count": 54,
623 |      "metadata": {},
624 |      "output_type": "execute_result"
625 |     }
626 |    ],
627 |    "source": [
628 |     "blend_df.corr()"
629 |    ]
630 |   },
631 |   {
632 |    "cell_type": "code",
633 |    "execution_count": 55,
634 |    "metadata": {
635 |     "collapsed": true
636 |    },
637 |    "outputs": [],
638 |    "source": [
639 |     "SUB_FILE_NAME = 'model_12_blend.csv'"
640 |    ]
641 |   },
642 |   {
643 |    "cell_type": "code",
644 |    "execution_count": 57,
645 |    "metadata": {},
646 |    "outputs": [
647 |     {
648 |      "data": {
649 |       "text/html": [
650 |        "<div>\n",
651 |        "<style scoped>\n",
652 |        "    .dataframe tbody tr th:only-of-type {\n",
653 |        "        vertical-align: middle;\n",
654 |        "    }\n",
655 |        "\n",
656 |        "    .dataframe tbody tr th {\n",
657 |        "        vertical-align: top;\n",
658 |        "    }\n",
659 |        "\n",
660 |        "    .dataframe thead th {\n",
661 |        "        text-align: right;\n",
662 |        "    }\n",
663 |        "</style>\n",
664 |        "<table border=\"1\" class=\"dataframe\">\n",
665 |        "  <thead>\n",
666 |        "    <tr style=\"text-align: right;\">\n",
667 |        "      <th></th>\n",
668 |        "      <th>ID</th>\n",
669 |        "      <th>target</th>\n",
670 |        "    </tr>\n",
671 |        "  </thead>\n",
672 |        "  <tbody>\n",
673 |        "    <tr>\n",
674 |        "      <th>0</th>\n",
675 |        "      <td>ID_test_0</td>\n",
676 |        "      <td>158.123774</td>\n",
677 |        "    </tr>\n",
678 |        "    <tr>\n",
679 |        "      <th>1</th>\n",
680 |        "      <td>ID_test_1</td>\n",
681 |        "      <td>97.217908</td>\n",
682 |        "    </tr>\n",
683 |        "    <tr>\n",
684 |        "      <th>2</th>\n",
685 |        "      <td>ID_test_10</td>\n",
686 |        "      <td>21.393733</td>\n",
687 |        "    </tr>\n",
688 |        "    <tr>\n",
689 |        "      <th>3</th>\n",
690 |        "      <td>ID_test_100</td>\n",
691 |        "      <td>63.222891</td>\n",
692 |        "    </tr>\n",
693 |        "    <tr>\n",
694 |        "      <th>4</th>\n",
695 |        "      <td>ID_test_1000</td>\n",
696 |        "      <td>92.046200</td>\n",
697 |        "    </tr>\n",
698 |        "    <tr>\n",
699 |        "      <th>5</th>\n",
700 |        "      <td>ID_test_1001</td>\n",
701 |        "      <td>44.955298</td>\n",
702 |        "    </tr>\n",
703 |        "    <tr>\n",
704 |        "      <th>6</th>\n",
705 |        "      <td>ID_test_1002</td>\n",
706 |        "      <td>83.270765</td>\n",
707 |        "    </tr>\n",
708 |        "    <tr>\n",
709 |        "      <th>7</th>\n",
710 |        "      <td>ID_test_1003</td>\n",
711 |        "      <td>36.458014</td>\n",
712 |        "    </tr>\n",
713 |        "    <tr>\n",
714 |        "      <th>8</th>\n",
715 |        "      <td>ID_test_1004</td>\n",
716 |        "      <td>34.101068</td>\n",
717 |        "    </tr>\n",
718 |        "    <tr>\n",
719 |        "      <th>9</th>\n",
720 |        "      <td>ID_test_1005</td>\n",
721 |        "      <td>47.728921</td>\n",
722 |        "    </tr>\n",
723 |        "  </tbody>\n",
724 |        "</table>\n",
725 |        "</div>"
726 |       ],
727 |       "text/plain": [
728 |        "             ID      target\n",
729 |        "0     ID_test_0  158.123774\n",
730 |        "1     ID_test_1   97.217908\n",
731 |        "2    ID_test_10   21.393733\n",
732 |        "3   ID_test_100   63.222891\n",
733 |        "4  ID_test_1000   92.046200\n",
734 |        "5  ID_test_1001   44.955298\n",
735 |        "6  ID_test_1002   83.270765\n",
736 |        "7  ID_test_1003   36.458014\n",
737 |        "8  ID_test_1004   34.101068\n",
738 |        "9  ID_test_1005   47.728921"
739 |       ]
740 |      },
741 |      "execution_count": 57,
742 |      "metadata": {},
743 |      "output_type": "execute_result"
744 |     }
745 |    ],
746 |    "source": [
747 |     "blend_df[['ID', 'target']].head(10)"
748 |    ]
749 |   },
750 |   {
751 |    "cell_type": "code",
752 |    "execution_count": 56,
753 |    "metadata": {},
754 |    "outputs": [
755 |     {
756 |      "data": {
757 |       "text/html": [
758 |        "<a href=model_12_blend.csv>Download CSV file</a>"
759 |       ],
760 |       "text/plain": [
761 |        "<IPython.core.display.HTML object>"
762 |       ]
763 |      },
764 |      "execution_count": 56,
765 |      "metadata": {},
766 |      "output_type": "execute_result"
767 |     }
768 |    ],
769 |    "source": [
770 |     "from IPython.display import HTML\n",
771 |     "def create_download_link(title = \"Download CSV file\", filename = \"data.csv\"):  \n",
772 |     "    html = '<a href={filename}>{title}</a>'\n",
773 |     "    html = html.format(title=title,filename=filename)\n",
774 |     "    return HTML(html)\n",
775 |     "create_download_link(filename = SUB_FILE_NAME)"
776 |    ]
777 |   }
778 |  ],
779 |  "metadata": {
780 |   "kernelspec": {
781 |    "display_name": "Python 3",
782 |    "language": "python",
783 |    "name": "python3"
784 |   },
785 |   "language_info": {
786 |    "codemirror_mode": {
787 |     "name": "ipython",
788 |     "version": 3
789 |    },
790 |    "file_extension": ".py",
791 |    "mimetype": "text/x-python",
792 |    "name": "python",
793 |    "nbconvert_exporter": "python",
794 |    "pygments_lexer": "ipython3",
795 |    "version": "3.6.3"
796 |   }
797 |  },
798 |  "nbformat": 4,
799 |  "nbformat_minor": 4
800 | }
801 | 


--------------------------------------------------------------------------------
/darius-model-2.ipynb:
--------------------------------------------------------------------------------
   1 | {
   2 |  "cells": [
   3 |   {
   4 |    "cell_type": "code",
   5 |    "execution_count": 25,
   6 |    "metadata": {
   7 |     "id": "pk5WiopT9oJw",
   8 |     "outputId": "f81ff94a-3399-4a56-9365-6c5e54d80782"
   9 |    },
  10 |    "outputs": [
  11 |     {
  12 |      "name": "stdout",
  13 |      "output_type": "stream",
  14 |      "text": [
  15 |       "\u001b[33mWARNING: You are using pip version 20.1; however, version 20.1.1 is available.\r\n",
  16 |       "You should consider upgrading via the '/opt/conda/bin/python3.7 -m pip install --upgrade pip' command.\u001b[0m\r\n"
  17 |      ]
  18 |     }
  19 |    ],
  20 |    "source": [
  21 |     "# installing catboost\n",
  22 |     "# Catboost == 0.22 was the version of catboost at the start of this competition\n",
  23 |     "!pip install catboost==0.22 --quiet"
  24 |    ]
  25 |   },
  26 |   {
  27 |    "cell_type": "code",
  28 |    "execution_count": 27,
  29 |    "metadata": {
  30 |     "collapsed": true,
  31 |     "id": "jVXG4QlTuVYr"
  32 |    },
  33 |    "outputs": [],
  34 |    "source": [
  35 |     "# Importing libraries\n",
  36 |     "import pandas as pd\n",
  37 |     "import numpy as np\n",
  38 |     "from lightgbm import LGBMRegressor\n",
  39 |     "from xgboost import XGBRegressor, XGBRFRegressor\n",
  40 |     "from sklearn.model_selection import KFold, cross_val_score, train_test_split\n",
  41 |     "from sklearn.metrics import mean_squared_error\n",
  42 |     "from sklearn.utils import shuffle\n",
  43 |     "from tqdm import tqdm, tqdm_notebook\n",
  44 |     "from functools import reduce\n",
  45 |     "from catboost import CatBoostRegressor, CatBoostClassifier\n",
  46 |     "import joblib\n",
  47 |     "\n",
  48 |     "\n",
  49 |     "import warnings\n",
  50 |     "warnings.filterwarnings('ignore')\n"
  51 |    ]
  52 |   },
  53 |   {
  54 |    "cell_type": "code",
  55 |    "execution_count": 28,
  56 |    "metadata": {
  57 |     "collapsed": true,
  58 |     "id": "agGrUcYwQ7on",
  59 |     "outputId": "5d4d760c-135f-4e6e-ae14-bc275b174734"
  60 |    },
  61 |    "outputs": [],
  62 |    "source": [
  63 |     "# Loading data\n",
  64 |     "train = pd.read_csv('./input/Train.csv')\n",
  65 |     "test = pd.read_csv('./input/Test.csv')"
  66 |    ]
  67 |   },
  68 |   {
  69 |    "cell_type": "code",
  70 |    "execution_count": 29,
  71 |    "metadata": {
  72 |     "collapsed": true,
  73 |     "id": "jNB_PDTNpxu4"
  74 |    },
  75 |    "outputs": [],
  76 |    "source": [
  77 |     "# Feature interaction functions\n",
  78 |     "# There are 4 types of interactions: product interactions, division interactions, sum interactions and divide interactions\n",
  79 |     "\n",
  80 |     "def add_prod_interacts(df, inter_cols):   \n",
  81 |     "    def apply_interacts(x, inter_cols):\n",
  82 |     "        cols = [x + '_prod_' + c for c in inter_cols[inter_cols.index(x)+1:]]\n",
  83 |     "        interacts_df[cols] = pd.concat([df[x] * df[c] for c in inter_cols[inter_cols.index(x)+1:]], axis=1)\n",
  84 |     "    \n",
  85 |     "    interacts_df = pd.DataFrame()\n",
  86 |     "    _ = df[inter_cols[:-1]].apply(lambda x: apply_interacts(x.name, inter_cols))\n",
  87 |     "    df = pd.concat([df, interacts_df], axis=1)\n",
  88 |     "    return df\n",
  89 |     "\n",
  90 |     "\n",
  91 |     "def add_div_interacts(df, inter_cols):   \n",
  92 |     "    def apply_interacts(x, inter_cols):\n",
  93 |     "        cols = [x + '_div_' + c for c in inter_cols[inter_cols.index(x)+1:]]\n",
  94 |     "        interacts_df[cols] = pd.concat([df[x] / df[c] for c in inter_cols[inter_cols.index(x)+1:]], axis=1)\n",
  95 |     "    \n",
  96 |     "    interacts_df = pd.DataFrame()\n",
  97 |     "    _ = df[inter_cols[:-1]].apply(lambda x: apply_interacts(x.name, inter_cols))\n",
  98 |     "    df = pd.concat([df, interacts_df], axis=1)\n",
  99 |     "\n",
 100 |     "    return df\n",
 101 |     "\n",
 102 |     "def add_sum_interacts(df, inter_cols):   \n",
 103 |     "    def apply_interacts(x, inter_cols):\n",
 104 |     "        cols = [x + '_sum_' + c for c in inter_cols[inter_cols.index(x)+1:]]\n",
 105 |     "        interacts_df[cols] = pd.concat([df[x] + df[c] for c in inter_cols[inter_cols.index(x)+1:]], axis=1)\n",
 106 |     "    \n",
 107 |     "    interacts_df = pd.DataFrame()\n",
 108 |     "    _ = df[inter_cols[:-1]].apply(lambda x: apply_interacts(x.name, inter_cols))\n",
 109 |     "    df = pd.concat([df, interacts_df], axis=1)\n",
 110 |     "\n",
 111 |     "    return df\n",
 112 |     "\n",
 113 |     "def add_diff_interacts(df, inter_cols):   \n",
 114 |     "    def apply_interacts(x, inter_cols):\n",
 115 |     "        cols = [x + '_diff_' + c for c in inter_cols[inter_cols.index(x)+1:]]\n",
 116 |     "        interacts_df[cols] = pd.concat([df[x] - df[c] for c in inter_cols[inter_cols.index(x)+1:]], axis=1)\n",
 117 |     "    \n",
 118 |     "    interacts_df = pd.DataFrame()\n",
 119 |     "    _ = df[inter_cols[:-1]].apply(lambda x: apply_interacts(x.name, inter_cols))\n",
 120 |     "    df = pd.concat([df, interacts_df], axis=1)\n",
 121 |     "\n",
 122 |     "    return df"
 123 |    ]
 124 |   },
 125 |   {
 126 |    "cell_type": "code",
 127 |    "execution_count": 30,
 128 |    "metadata": {
 129 |     "collapsed": true,
 130 |     "id": "VJjtuM3kvCAT"
 131 |    },
 132 |    "outputs": [],
 133 |    "source": [
 134 |     "# Loading data\n",
 135 |     "train = pd.read_csv('./input/Train.csv')\n",
 136 |     "test = pd.read_csv('./input/Test.csv')"
 137 |    ]
 138 |   },
 139 |   {
 140 |    "cell_type": "code",
 141 |    "execution_count": 31,
 142 |    "metadata": {
 143 |     "collapsed": true,
 144 |     "id": "4MhDKOpYF9dm"
 145 |    },
 146 |    "outputs": [],
 147 |    "source": [
 148 |     "# Separating the target variable from the training dataframe\n",
 149 |     "#\n",
 150 |     "target = train.target\n",
 151 |     "\n",
 152 |     "# Aligning the train and test dataframes\n",
 153 |     "#\n",
 154 |     "train, test = train.align(test, join = 'inner',axis = 1)\n",
 155 |     "\n",
 156 |     "# Creating a separator column to both train and test,\n",
 157 |     "# This is to be used in separation\n",
 158 |     "#\n",
 159 |     "train['separator'] = 0\n",
 160 |     "test['separator'] = 1\n",
 161 |     "# Combing the train and test dataframes together\n",
 162 |     "#\n",
 163 |     "comb = pd.concat([train, test])\n",
 164 |     "\n",
 165 |     "# Creating a function to replace all spaces in the dataframe with np.nan\n",
 166 |     "#\n",
 167 |     "def replace_nan(x):\n",
 168 |     "    if x==\" \":\n",
 169 |     "        return np.nan\n",
 170 |     "    else :\n",
 171 |     "        return float(x)\n",
 172 |     "\n",
 173 |     "# Creating a list of the main columns\n",
 174 |     "#\n",
 175 |     "main_cols = [\"temp\",\"precip\",\"rel_humidity\",\"wind_dir\",\"wind_spd\",\"atmos_press\"]\n",
 176 |     "\n",
 177 |     "# Replacing spaces with np.nan\n",
 178 |     "#\n",
 179 |     "for col in main_cols: \n",
 180 |     "    comb[col]=comb[col].apply(lambda x: [ replace_nan(X) for X in x.replace(\"nan\",\" \").split(\",\")])\n",
 181 |     "\n",
 182 |     "def make_columns(feature):\n",
 183 |     "    return [f\"{feature}_{i}\" for i in range(1, 122)]\n",
 184 |     "    \n",
 185 |     "# Generating dataframes of hours for each main column\n",
 186 |     "#\n",
 187 |     "comb_temp         = pd.DataFrame([x for x in comb.temp],         columns=make_columns('temp'))\n",
 188 |     "comb_precip       = pd.DataFrame([x for x in comb.precip],       columns=make_columns('precip'))\n",
 189 |     "comb_rel_humidity = pd.DataFrame([x for x in comb.rel_humidity], columns=make_columns('rel_humidity'))\n",
 190 |     "comb_wind_dir     = pd.DataFrame([x for x in comb.wind_dir],     columns=make_columns('wind_dir'))\n",
 191 |     "comb_wind_spd     = pd.DataFrame([x for x in comb.wind_spd],     columns=make_columns('wind_spd'))\n",
 192 |     "comb_atmos_press  = pd.DataFrame([x for x in comb.atmos_press],  columns=make_columns('atmos_press'))\n",
 193 |     "\n",
 194 |     "comb_temp['ID'], comb_precip['ID'], comb_rel_humidity['ID'], comb_wind_dir['ID'], comb_wind_spd['ID'], comb_atmos_press['ID'] = [list(comb.ID)] * 6\n",
 195 |     "\n",
 196 |     "# Combining the generated dataframes together\n",
 197 |     "#\n",
 198 |     "comb_dfs = [comb, comb_temp, comb_precip, comb_rel_humidity, comb_wind_dir, comb_wind_spd, comb_atmos_press]\n",
 199 |     "comb = reduce(lambda  left,right: pd.merge(left,right,on=['ID'], how='outer'), comb_dfs)\n",
 200 |     "comb.drop(main_cols, axis = 1, inplace = True)\n",
 201 |     "df = comb.copy()"
 202 |    ]
 203 |   },
 204 |   {
 205 |    "cell_type": "code",
 206 |    "execution_count": 32,
 207 |    "metadata": {
 208 |     "collapsed": true,
 209 |     "id": "pEyFVf1lqInY"
 210 |    },
 211 |    "outputs": [],
 212 |    "source": [
 213 |     "comb = df.copy()"
 214 |    ]
 215 |   },
 216 |   {
 217 |    "cell_type": "code",
 218 |    "execution_count": 33,
 219 |    "metadata": {
 220 |     "id": "xzn_w6FYftB8",
 221 |     "outputId": "9c6b5360-766f-4c1b-c223-c00ceaa834f9"
 222 |    },
 223 |    "outputs": [
 224 |     {
 225 |      "data": {
 226 |       "application/vnd.jupyter.widget-view+json": {
 227 |        "model_id": "999f9adfcfd444af978f061a15a74ea8",
 228 |        "version_major": 2,
 229 |        "version_minor": 0
 230 |       },
 231 |       "text/plain": [
 232 |        "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
 233 |       ]
 234 |      },
 235 |      "metadata": {},
 236 |      "output_type": "display_data"
 237 |     },
 238 |     {
 239 |      "data": {
 240 |       "application/vnd.jupyter.widget-view+json": {
 241 |        "model_id": "",
 242 |        "version_major": 2,
 243 |        "version_minor": 0
 244 |       },
 245 |       "text/plain": [
 246 |        "HBox(children=(FloatProgress(value=0.0, max=11.0), HTML(value='')))"
 247 |       ]
 248 |      },
 249 |      "metadata": {},
 250 |      "output_type": "display_data"
 251 |     },
 252 |     {
 253 |      "data": {
 254 |       "application/vnd.jupyter.widget-view+json": {
 255 |        "model_id": "",
 256 |        "version_major": 2,
 257 |        "version_minor": 0
 258 |       },
 259 |       "text/plain": [
 260 |        "HBox(children=(FloatProgress(value=0.0, max=11.0), HTML(value='')))"
 261 |       ]
 262 |      },
 263 |      "metadata": {},
 264 |      "output_type": "display_data"
 265 |     },
 266 |     {
 267 |      "data": {
 268 |       "application/vnd.jupyter.widget-view+json": {
 269 |        "model_id": "",
 270 |        "version_major": 2,
 271 |        "version_minor": 0
 272 |       },
 273 |       "text/plain": [
 274 |        "HBox(children=(FloatProgress(value=0.0, max=11.0), HTML(value='')))"
 275 |       ]
 276 |      },
 277 |      "metadata": {},
 278 |      "output_type": "display_data"
 279 |     },
 280 |     {
 281 |      "data": {
 282 |       "application/vnd.jupyter.widget-view+json": {
 283 |        "model_id": "",
 284 |        "version_major": 2,
 285 |        "version_minor": 0
 286 |       },
 287 |       "text/plain": [
 288 |        "HBox(children=(FloatProgress(value=0.0, max=11.0), HTML(value='')))"
 289 |       ]
 290 |      },
 291 |      "metadata": {},
 292 |      "output_type": "display_data"
 293 |     },
 294 |     {
 295 |      "data": {
 296 |       "application/vnd.jupyter.widget-view+json": {
 297 |        "model_id": "",
 298 |        "version_major": 2,
 299 |        "version_minor": 0
 300 |       },
 301 |       "text/plain": [
 302 |        "HBox(children=(FloatProgress(value=0.0, max=11.0), HTML(value='')))"
 303 |       ]
 304 |      },
 305 |      "metadata": {},
 306 |      "output_type": "display_data"
 307 |     },
 308 |     {
 309 |      "data": {
 310 |       "application/vnd.jupyter.widget-view+json": {
 311 |        "model_id": "",
 312 |        "version_major": 2,
 313 |        "version_minor": 0
 314 |       },
 315 |       "text/plain": [
 316 |        "HBox(children=(FloatProgress(value=0.0, max=11.0), HTML(value='')))"
 317 |       ]
 318 |      },
 319 |      "metadata": {},
 320 |      "output_type": "display_data"
 321 |     },
 322 |     {
 323 |      "name": "stdout",
 324 |      "output_type": "stream",
 325 |      "text": [
 326 |       "\n"
 327 |      ]
 328 |     }
 329 |    ],
 330 |    "source": [
 331 |     "# Adding aggregation features for each variable\n",
 332 |     "#\n",
 333 |     "aggs = ['mean', 'std', 'var', 'kurt', 'skew', 'max', 'median', 'sum', 'mode', 'sem', 'min']\n",
 334 |     "for col in tqdm_notebook(main_cols):\n",
 335 |     "  for ag in tqdm_notebook(aggs, leave = False):\n",
 336 |     "    if ag == 'mode':\n",
 337 |     "      comb[col[0] + col[-1] + '_'+ag] = comb[[x for x in comb.columns if x.startswith(col)]].agg(ag, axis = 1)[0]\n",
 338 |     "    else:\n",
 339 |     "      comb[col[0] + col[-1] + '_'+ag] = comb[[x for x in comb.columns if x.startswith(col)]].agg(ag, axis = 1)\n",
 340 |     "\n",
 341 |     "# Creating separate dataframes for each variable\n",
 342 |     "# Creating a list of columns for each separate dataframe\n",
 343 |     "#\n",
 344 |     "temp, temp_cols  = comb[[x for x in comb.columns if x.startswith('temp')]], [x for x in comb.columns if x.startswith('temp')]\n",
 345 |     "precip, precip_cols = comb[[x for x in comb.columns if x.startswith('precip')]], [x for x in comb.columns if x.startswith('precip')]\n",
 346 |     "humid, humid_cols = comb[[x for x in comb.columns if x.startswith('rel_humidity')]], [x for x in comb.columns if x.startswith('rel_humidity')]\n",
 347 |     "wind_dir, wind_dir_cols = comb[[x for x in comb.columns if x.startswith('wind_dir')]], [x for x in comb.columns if x.startswith('wind_dir')]\n",
 348 |     "wind_spd, wind_spd_cols = comb[[x for x in comb.columns if x.startswith('wind_spd')]], [x for x in comb.columns if x.startswith('wind_spd')]\n",
 349 |     "atmp, atmp_cols = comb[[x for x in comb.columns if x.startswith('atmos_press')]], [x for x in comb.columns if x.startswith('atmos_press')]\n",
 350 |     "fill_cols = comb.columns"
 351 |    ]
 352 |   },
 353 |   {
 354 |    "cell_type": "code",
 355 |    "execution_count": 34,
 356 |    "metadata": {
 357 |     "id": "XHCE-yW1r3q8",
 358 |     "outputId": "dd3d6870-4a06-4794-c94c-eba320fa9335"
 359 |    },
 360 |    "outputs": [
 361 |     {
 362 |      "data": {
 363 |       "text/html": [
 364 |        "<div>\n",
 365 |        "<style scoped>\n",
 366 |        "    .dataframe tbody tr th:only-of-type {\n",
 367 |        "        vertical-align: middle;\n",
 368 |        "    }\n",
 369 |        "\n",
 370 |        "    .dataframe tbody tr th {\n",
 371 |        "        vertical-align: top;\n",
 372 |        "    }\n",
 373 |        "\n",
 374 |        "    .dataframe thead th {\n",
 375 |        "        text-align: right;\n",
 376 |        "    }\n",
 377 |        "</style>\n",
 378 |        "<table border=\"1\" class=\"dataframe\">\n",
 379 |        "  <thead>\n",
 380 |        "    <tr style=\"text-align: right;\">\n",
 381 |        "      <th></th>\n",
 382 |        "      <th>ID</th>\n",
 383 |        "      <th>location</th>\n",
 384 |        "      <th>separator</th>\n",
 385 |        "      <th>temp_1</th>\n",
 386 |        "      <th>temp_2</th>\n",
 387 |        "      <th>temp_3</th>\n",
 388 |        "      <th>temp_4</th>\n",
 389 |        "      <th>temp_5</th>\n",
 390 |        "      <th>temp_6</th>\n",
 391 |        "      <th>temp_7</th>\n",
 392 |        "      <th>...</th>\n",
 393 |        "      <th>as_std</th>\n",
 394 |        "      <th>as_var</th>\n",
 395 |        "      <th>as_kurt</th>\n",
 396 |        "      <th>as_skew</th>\n",
 397 |        "      <th>as_max</th>\n",
 398 |        "      <th>as_median</th>\n",
 399 |        "      <th>as_sum</th>\n",
 400 |        "      <th>as_mode</th>\n",
 401 |        "      <th>as_sem</th>\n",
 402 |        "      <th>as_min</th>\n",
 403 |        "    </tr>\n",
 404 |        "  </thead>\n",
 405 |        "  <tbody>\n",
 406 |        "    <tr>\n",
 407 |        "      <th>0</th>\n",
 408 |        "      <td>ID_train_0</td>\n",
 409 |        "      <td>C</td>\n",
 410 |        "      <td>0</td>\n",
 411 |        "      <td>NaN</td>\n",
 412 |        "      <td>NaN</td>\n",
 413 |        "      <td>NaN</td>\n",
 414 |        "      <td>NaN</td>\n",
 415 |        "      <td>NaN</td>\n",
 416 |        "      <td>NaN</td>\n",
 417 |        "      <td>NaN</td>\n",
 418 |        "      <td>...</td>\n",
 419 |        "      <td>0.072682</td>\n",
 420 |        "      <td>0.005283</td>\n",
 421 |        "      <td>-0.158696</td>\n",
 422 |        "      <td>-0.383144</td>\n",
 423 |        "      <td>87.871667</td>\n",
 424 |        "      <td>87.762083</td>\n",
 425 |        "      <td>1404.038939</td>\n",
 426 |        "      <td>87.614167</td>\n",
 427 |        "      <td>0.018170</td>\n",
 428 |        "      <td>87.614167</td>\n",
 429 |        "    </tr>\n",
 430 |        "    <tr>\n",
 431 |        "      <th>1</th>\n",
 432 |        "      <td>ID_train_1</td>\n",
 433 |        "      <td>D</td>\n",
 434 |        "      <td>0</td>\n",
 435 |        "      <td>22.533333</td>\n",
 436 |        "      <td>21.716667</td>\n",
 437 |        "      <td>20.833333</td>\n",
 438 |        "      <td>20.983333</td>\n",
 439 |        "      <td>20.875000</td>\n",
 440 |        "      <td>20.141667</td>\n",
 441 |        "      <td>19.375000</td>\n",
 442 |        "      <td>...</td>\n",
 443 |        "      <td>0.156648</td>\n",
 444 |        "      <td>0.024539</td>\n",
 445 |        "      <td>-0.446340</td>\n",
 446 |        "      <td>-0.173356</td>\n",
 447 |        "      <td>90.725000</td>\n",
 448 |        "      <td>90.429167</td>\n",
 449 |        "      <td>10942.020833</td>\n",
 450 |        "      <td>90.219167</td>\n",
 451 |        "      <td>0.014241</td>\n",
 452 |        "      <td>90.056667</td>\n",
 453 |        "    </tr>\n",
 454 |        "    <tr>\n",
 455 |        "      <th>2</th>\n",
 456 |        "      <td>ID_train_10</td>\n",
 457 |        "      <td>A</td>\n",
 458 |        "      <td>0</td>\n",
 459 |        "      <td>28.975000</td>\n",
 460 |        "      <td>27.950000</td>\n",
 461 |        "      <td>29.600000</td>\n",
 462 |        "      <td>26.425000</td>\n",
 463 |        "      <td>22.091667</td>\n",
 464 |        "      <td>21.775000</td>\n",
 465 |        "      <td>22.333333</td>\n",
 466 |        "      <td>...</td>\n",
 467 |        "      <td>0.180233</td>\n",
 468 |        "      <td>0.032484</td>\n",
 469 |        "      <td>-0.227481</td>\n",
 470 |        "      <td>-0.243561</td>\n",
 471 |        "      <td>88.813333</td>\n",
 472 |        "      <td>88.425000</td>\n",
 473 |        "      <td>10610.511667</td>\n",
 474 |        "      <td>88.287500</td>\n",
 475 |        "      <td>0.016453</td>\n",
 476 |        "      <td>87.982500</td>\n",
 477 |        "    </tr>\n",
 478 |        "    <tr>\n",
 479 |        "      <th>3</th>\n",
 480 |        "      <td>ID_train_100</td>\n",
 481 |        "      <td>A</td>\n",
 482 |        "      <td>0</td>\n",
 483 |        "      <td>22.966667</td>\n",
 484 |        "      <td>24.266667</td>\n",
 485 |        "      <td>25.275000</td>\n",
 486 |        "      <td>25.625000</td>\n",
 487 |        "      <td>25.866667</td>\n",
 488 |        "      <td>25.091667</td>\n",
 489 |        "      <td>24.025000</td>\n",
 490 |        "      <td>...</td>\n",
 491 |        "      <td>0.162430</td>\n",
 492 |        "      <td>0.026384</td>\n",
 493 |        "      <td>-0.462889</td>\n",
 494 |        "      <td>-0.347610</td>\n",
 495 |        "      <td>88.685000</td>\n",
 496 |        "      <td>88.400000</td>\n",
 497 |        "      <td>10693.606667</td>\n",
 498 |        "      <td>88.271667</td>\n",
 499 |        "      <td>0.014766</td>\n",
 500 |        "      <td>87.965000</td>\n",
 501 |        "    </tr>\n",
 502 |        "    <tr>\n",
 503 |        "      <th>4</th>\n",
 504 |        "      <td>ID_train_1000</td>\n",
 505 |        "      <td>A</td>\n",
 506 |        "      <td>0</td>\n",
 507 |        "      <td>21.875000</td>\n",
 508 |        "      <td>21.575000</td>\n",
 509 |        "      <td>21.525000</td>\n",
 510 |        "      <td>21.433333</td>\n",
 511 |        "      <td>20.508333</td>\n",
 512 |        "      <td>19.916667</td>\n",
 513 |        "      <td>18.991667</td>\n",
 514 |        "      <td>...</td>\n",
 515 |        "      <td>0.120393</td>\n",
 516 |        "      <td>0.014494</td>\n",
 517 |        "      <td>-0.062557</td>\n",
 518 |        "      <td>-0.705667</td>\n",
 519 |        "      <td>88.719167</td>\n",
 520 |        "      <td>88.552500</td>\n",
 521 |        "      <td>2656.143106</td>\n",
 522 |        "      <td>88.268333</td>\n",
 523 |        "      <td>0.021981</td>\n",
 524 |        "      <td>88.268333</td>\n",
 525 |        "    </tr>\n",
 526 |        "  </tbody>\n",
 527 |        "</table>\n",
 528 |        "<p>5 rows × 795 columns</p>\n",
 529 |        "</div>"
 530 |       ],
 531 |       "text/plain": [
 532 |        "              ID location  separator     temp_1     temp_2     temp_3  \\\n",
 533 |        "0     ID_train_0        C          0        NaN        NaN        NaN   \n",
 534 |        "1     ID_train_1        D          0  22.533333  21.716667  20.833333   \n",
 535 |        "2    ID_train_10        A          0  28.975000  27.950000  29.600000   \n",
 536 |        "3   ID_train_100        A          0  22.966667  24.266667  25.275000   \n",
 537 |        "4  ID_train_1000        A          0  21.875000  21.575000  21.525000   \n",
 538 |        "\n",
 539 |        "      temp_4     temp_5     temp_6     temp_7  ...    as_std    as_var  \\\n",
 540 |        "0        NaN        NaN        NaN        NaN  ...  0.072682  0.005283   \n",
 541 |        "1  20.983333  20.875000  20.141667  19.375000  ...  0.156648  0.024539   \n",
 542 |        "2  26.425000  22.091667  21.775000  22.333333  ...  0.180233  0.032484   \n",
 543 |        "3  25.625000  25.866667  25.091667  24.025000  ...  0.162430  0.026384   \n",
 544 |        "4  21.433333  20.508333  19.916667  18.991667  ...  0.120393  0.014494   \n",
 545 |        "\n",
 546 |        "    as_kurt   as_skew     as_max  as_median        as_sum    as_mode  \\\n",
 547 |        "0 -0.158696 -0.383144  87.871667  87.762083   1404.038939  87.614167   \n",
 548 |        "1 -0.446340 -0.173356  90.725000  90.429167  10942.020833  90.219167   \n",
 549 |        "2 -0.227481 -0.243561  88.813333  88.425000  10610.511667  88.287500   \n",
 550 |        "3 -0.462889 -0.347610  88.685000  88.400000  10693.606667  88.271667   \n",
 551 |        "4 -0.062557 -0.705667  88.719167  88.552500   2656.143106  88.268333   \n",
 552 |        "\n",
 553 |        "     as_sem     as_min  \n",
 554 |        "0  0.018170  87.614167  \n",
 555 |        "1  0.014241  90.056667  \n",
 556 |        "2  0.016453  87.982500  \n",
 557 |        "3  0.014766  87.965000  \n",
 558 |        "4  0.021981  88.268333  \n",
 559 |        "\n",
 560 |        "[5 rows x 795 columns]"
 561 |       ]
 562 |      },
 563 |      "execution_count": 34,
 564 |      "metadata": {},
 565 |      "output_type": "execute_result"
 566 |     }
 567 |    ],
 568 |    "source": [
 569 |     "# Previewing the head of the generated dataframe\n",
 570 |     "#\n",
 571 |     "comb.head()"
 572 |    ]
 573 |   },
 574 |   {
 575 |    "cell_type": "code",
 576 |    "execution_count": 35,
 577 |    "metadata": {
 578 |     "id": "v4DlbAzKp_c8",
 579 |     "outputId": "26c4e9d8-68ae-4ed1-8e9a-e9a009f7343c"
 580 |    },
 581 |    "outputs": [
 582 |     {
 583 |      "data": {
 584 |       "application/vnd.jupyter.widget-view+json": {
 585 |        "model_id": "27712189b04b46b68969a0482c5243a5",
 586 |        "version_major": 2,
 587 |        "version_minor": 0
 588 |       },
 589 |       "text/plain": [
 590 |        "HBox(children=(FloatProgress(value=0.0, max=795.0), HTML(value='')))"
 591 |       ]
 592 |      },
 593 |      "metadata": {},
 594 |      "output_type": "display_data"
 595 |     },
 596 |     {
 597 |      "name": "stdout",
 598 |      "output_type": "stream",
 599 |      "text": [
 600 |       "\n"
 601 |      ]
 602 |     },
 603 |     {
 604 |      "data": {
 605 |       "application/vnd.jupyter.widget-view+json": {
 606 |        "model_id": "",
 607 |        "version_major": 2,
 608 |        "version_minor": 0
 609 |       },
 610 |       "text/plain": [
 611 |        "HBox(children=(FloatProgress(value=0.0, max=3.0), HTML(value='')))"
 612 |       ]
 613 |      },
 614 |      "metadata": {},
 615 |      "output_type": "display_data"
 616 |     }
 617 |    ],
 618 |    "source": [
 619 |     "# Creating a list of columns containing aggregates only\n",
 620 |     "cols_mean = []\n",
 621 |     "cols_max = []\n",
 622 |     "cols_min = []\n",
 623 |     "for x in tqdm_notebook(comb.columns):\n",
 624 |     "  if 'mean' in x:\n",
 625 |     "    cols_mean.append(x)\n",
 626 |     "  elif 'max' in x:\n",
 627 |     "    cols_max.append(x)\n",
 628 |     "  elif 'min' in x:\n",
 629 |     "    cols_min.append(x)\n",
 630 |     "  else:\n",
 631 |     "    pass\n",
 632 |     "\n",
 633 |     "# Generating feature interactions between aggregates only\n",
 634 |     "for num_cols in tqdm_notebook([cols_mean, cols_max, cols_min], leave = False):\n",
 635 |     "    comb = add_prod_interacts(comb, num_cols)\n",
 636 |     "    comb = add_div_interacts(comb, num_cols)\n",
 637 |     "    comb = add_diff_interacts(comb, num_cols)"
 638 |    ]
 639 |   },
 640 |   {
 641 |    "cell_type": "code",
 642 |    "execution_count": 36,
 643 |    "metadata": {
 644 |     "collapsed": true,
 645 |     "id": "qdVrf3ZmcFZn"
 646 |    },
 647 |    "outputs": [],
 648 |    "source": [
 649 |     "# Generating new features, by adding each variable per hour\n",
 650 |     "for x, y, z, a, b in zip(temp.columns, precip.columns, humid.columns, wind_spd.columns, atmp.columns):\n",
 651 |     "  comb['add_tp' +y[-4:]] = temp[x] + precip[y] + humid[z] + wind_spd[a] + atmp[b]"
 652 |    ]
 653 |   },
 654 |   {
 655 |    "cell_type": "code",
 656 |    "execution_count": 37,
 657 |    "metadata": {
 658 |     "collapsed": true,
 659 |     "id": "IU2Z7uDigrgW"
 660 |    },
 661 |    "outputs": [],
 662 |    "source": [
 663 |     "# Filling missing values using forward fill\n",
 664 |     "comb = comb.ffill(axis = 1)"
 665 |    ]
 666 |   },
 667 |   {
 668 |    "cell_type": "code",
 669 |    "execution_count": 38,
 670 |    "metadata": {
 671 |     "collapsed": true,
 672 |     "id": "YnsfoPv_R5yN"
 673 |    },
 674 |    "outputs": [],
 675 |    "source": [
 676 |     "# Calculating the difference between features per each dataframe\n",
 677 |     "\n",
 678 |     "dfs = [temp, precip, humid, wind_spd, atmp]\n",
 679 |     "\n",
 680 |     "diff_dfs = []\n",
 681 |     "for i in range(5):\n",
 682 |     "  i = dfs[i]\n",
 683 |     "  name = str(i.columns[0].split('_')[0])\n",
 684 |     "  temp_df = i.diff(axis = 1).values\n",
 685 |     "  temp_df = pd.DataFrame(temp_df, columns=['diff_' +name + '_' + str(i) for i in range(1, 122)])\n",
 686 |     "  diff_dfs.append(temp_df)\n",
 687 |     "\n",
 688 |     "diff_dfs.append(comb)\n",
 689 |     "comb = reduce(lambda  left,right: pd.merge(left,right, right_index=True, left_index=True, how='outer'), diff_dfs)"
 690 |    ]
 691 |   },
 692 |   {
 693 |    "cell_type": "code",
 694 |    "execution_count": 39,
 695 |    "metadata": {
 696 |     "id": "TUeOD_AMUMyQ",
 697 |     "outputId": "ebd84e15-b073-45aa-d096-c2f270ce2e10"
 698 |    },
 699 |    "outputs": [
 700 |     {
 701 |      "data": {
 702 |       "text/html": [
 703 |        "<div>\n",
 704 |        "<style scoped>\n",
 705 |        "    .dataframe tbody tr th:only-of-type {\n",
 706 |        "        vertical-align: middle;\n",
 707 |        "    }\n",
 708 |        "\n",
 709 |        "    .dataframe tbody tr th {\n",
 710 |        "        vertical-align: top;\n",
 711 |        "    }\n",
 712 |        "\n",
 713 |        "    .dataframe thead th {\n",
 714 |        "        text-align: right;\n",
 715 |        "    }\n",
 716 |        "</style>\n",
 717 |        "<table border=\"1\" class=\"dataframe\">\n",
 718 |        "  <thead>\n",
 719 |        "    <tr style=\"text-align: right;\">\n",
 720 |        "      <th></th>\n",
 721 |        "      <th>diff_temp_1</th>\n",
 722 |        "      <th>diff_temp_2</th>\n",
 723 |        "      <th>diff_temp_3</th>\n",
 724 |        "      <th>diff_temp_4</th>\n",
 725 |        "      <th>diff_temp_5</th>\n",
 726 |        "      <th>diff_temp_6</th>\n",
 727 |        "      <th>diff_temp_7</th>\n",
 728 |        "      <th>diff_temp_8</th>\n",
 729 |        "      <th>diff_temp_9</th>\n",
 730 |        "      <th>diff_temp_10</th>\n",
 731 |        "      <th>...</th>\n",
 732 |        "      <th>add_tp_112</th>\n",
 733 |        "      <th>add_tp_113</th>\n",
 734 |        "      <th>add_tp_114</th>\n",
 735 |        "      <th>add_tp_115</th>\n",
 736 |        "      <th>add_tp_116</th>\n",
 737 |        "      <th>add_tp_117</th>\n",
 738 |        "      <th>add_tp_118</th>\n",
 739 |        "      <th>add_tp_119</th>\n",
 740 |        "      <th>add_tp_120</th>\n",
 741 |        "      <th>add_tp_121</th>\n",
 742 |        "    </tr>\n",
 743 |        "  </thead>\n",
 744 |        "  <tbody>\n",
 745 |        "    <tr>\n",
 746 |        "      <th>0</th>\n",
 747 |        "      <td>NaN</td>\n",
 748 |        "      <td>NaN</td>\n",
 749 |        "      <td>NaN</td>\n",
 750 |        "      <td>NaN</td>\n",
 751 |        "      <td>NaN</td>\n",
 752 |        "      <td>NaN</td>\n",
 753 |        "      <td>NaN</td>\n",
 754 |        "      <td>NaN</td>\n",
 755 |        "      <td>NaN</td>\n",
 756 |        "      <td>NaN</td>\n",
 757 |        "      <td>...</td>\n",
 758 |        "      <td>111.009</td>\n",
 759 |        "      <td>110.396</td>\n",
 760 |        "      <td>109.991</td>\n",
 761 |        "      <td>110.031</td>\n",
 762 |        "      <td>110.085</td>\n",
 763 |        "      <td>109.618</td>\n",
 764 |        "      <td>109.161</td>\n",
 765 |        "      <td>108.627</td>\n",
 766 |        "      <td>108.839</td>\n",
 767 |        "      <td>109.081</td>\n",
 768 |        "    </tr>\n",
 769 |        "    <tr>\n",
 770 |        "      <th>1</th>\n",
 771 |        "      <td>NaN</td>\n",
 772 |        "      <td>-0.816667</td>\n",
 773 |        "      <td>-0.883333</td>\n",
 774 |        "      <td>0.150000</td>\n",
 775 |        "      <td>-0.108333</td>\n",
 776 |        "      <td>-0.733333</td>\n",
 777 |        "      <td>-0.766667</td>\n",
 778 |        "      <td>-0.583333</td>\n",
 779 |        "      <td>-0.016667</td>\n",
 780 |        "      <td>-0.133333</td>\n",
 781 |        "      <td>...</td>\n",
 782 |        "      <td>118.268</td>\n",
 783 |        "      <td>119.334</td>\n",
 784 |        "      <td>120.191</td>\n",
 785 |        "      <td>122.078</td>\n",
 786 |        "      <td>122.734</td>\n",
 787 |        "      <td>123.072</td>\n",
 788 |        "      <td>121.885</td>\n",
 789 |        "      <td>119.338</td>\n",
 790 |        "      <td>118.589</td>\n",
 791 |        "      <td>114.357</td>\n",
 792 |        "    </tr>\n",
 793 |        "    <tr>\n",
 794 |        "      <th>2</th>\n",
 795 |        "      <td>NaN</td>\n",
 796 |        "      <td>-1.025000</td>\n",
 797 |        "      <td>1.650000</td>\n",
 798 |        "      <td>-3.175000</td>\n",
 799 |        "      <td>-4.333333</td>\n",
 800 |        "      <td>-0.316667</td>\n",
 801 |        "      <td>0.558333</td>\n",
 802 |        "      <td>-0.383333</td>\n",
 803 |        "      <td>-1.508333</td>\n",
 804 |        "      <td>0.000000</td>\n",
 805 |        "      <td>...</td>\n",
 806 |        "      <td>111.332</td>\n",
 807 |        "      <td>110.389</td>\n",
 808 |        "      <td>110.174</td>\n",
 809 |        "      <td>110.092</td>\n",
 810 |        "      <td>110.746</td>\n",
 811 |        "      <td>113.401</td>\n",
 812 |        "      <td>116.527</td>\n",
 813 |        "      <td>118.449</td>\n",
 814 |        "      <td>119.659</td>\n",
 815 |        "      <td>120.692</td>\n",
 816 |        "    </tr>\n",
 817 |        "    <tr>\n",
 818 |        "      <th>3</th>\n",
 819 |        "      <td>NaN</td>\n",
 820 |        "      <td>1.300000</td>\n",
 821 |        "      <td>1.008333</td>\n",
 822 |        "      <td>0.350000</td>\n",
 823 |        "      <td>0.241667</td>\n",
 824 |        "      <td>-0.775000</td>\n",
 825 |        "      <td>-1.066667</td>\n",
 826 |        "      <td>-1.600000</td>\n",
 827 |        "      <td>-2.191667</td>\n",
 828 |        "      <td>-2.816667</td>\n",
 829 |        "      <td>...</td>\n",
 830 |        "      <td>109.088</td>\n",
 831 |        "      <td>108.214</td>\n",
 832 |        "      <td>108.176</td>\n",
 833 |        "      <td>107.477</td>\n",
 834 |        "      <td>107.332</td>\n",
 835 |        "      <td>108.305</td>\n",
 836 |        "      <td>110.405</td>\n",
 837 |        "      <td>112.255</td>\n",
 838 |        "      <td>114.158</td>\n",
 839 |        "      <td>115.68</td>\n",
 840 |        "    </tr>\n",
 841 |        "    <tr>\n",
 842 |        "      <th>4</th>\n",
 843 |        "      <td>NaN</td>\n",
 844 |        "      <td>-0.300000</td>\n",
 845 |        "      <td>-0.050000</td>\n",
 846 |        "      <td>-0.091667</td>\n",
 847 |        "      <td>-0.925000</td>\n",
 848 |        "      <td>-0.591667</td>\n",
 849 |        "      <td>-0.925000</td>\n",
 850 |        "      <td>-0.400000</td>\n",
 851 |        "      <td>-0.541667</td>\n",
 852 |        "      <td>-0.133333</td>\n",
 853 |        "      <td>...</td>\n",
 854 |        "      <td>110.649</td>\n",
 855 |        "      <td>110.649</td>\n",
 856 |        "      <td>110.649</td>\n",
 857 |        "      <td>110.649</td>\n",
 858 |        "      <td>110.649</td>\n",
 859 |        "      <td>110.649</td>\n",
 860 |        "      <td>110.649</td>\n",
 861 |        "      <td>110.649</td>\n",
 862 |        "      <td>110.649</td>\n",
 863 |        "      <td>110.649</td>\n",
 864 |        "    </tr>\n",
 865 |        "  </tbody>\n",
 866 |        "</table>\n",
 867 |        "<p>5 rows × 1656 columns</p>\n",
 868 |        "</div>"
 869 |       ],
 870 |       "text/plain": [
 871 |        "   diff_temp_1  diff_temp_2  diff_temp_3  diff_temp_4  diff_temp_5  \\\n",
 872 |        "0          NaN          NaN          NaN          NaN          NaN   \n",
 873 |        "1          NaN    -0.816667    -0.883333     0.150000    -0.108333   \n",
 874 |        "2          NaN    -1.025000     1.650000    -3.175000    -4.333333   \n",
 875 |        "3          NaN     1.300000     1.008333     0.350000     0.241667   \n",
 876 |        "4          NaN    -0.300000    -0.050000    -0.091667    -0.925000   \n",
 877 |        "\n",
 878 |        "   diff_temp_6  diff_temp_7  diff_temp_8  diff_temp_9  diff_temp_10  ...  \\\n",
 879 |        "0          NaN          NaN          NaN          NaN           NaN  ...   \n",
 880 |        "1    -0.733333    -0.766667    -0.583333    -0.016667     -0.133333  ...   \n",
 881 |        "2    -0.316667     0.558333    -0.383333    -1.508333      0.000000  ...   \n",
 882 |        "3    -0.775000    -1.066667    -1.600000    -2.191667     -2.816667  ...   \n",
 883 |        "4    -0.591667    -0.925000    -0.400000    -0.541667     -0.133333  ...   \n",
 884 |        "\n",
 885 |        "   add_tp_112  add_tp_113  add_tp_114  add_tp_115  add_tp_116  add_tp_117  \\\n",
 886 |        "0     111.009     110.396     109.991     110.031     110.085     109.618   \n",
 887 |        "1     118.268     119.334     120.191     122.078     122.734     123.072   \n",
 888 |        "2     111.332     110.389     110.174     110.092     110.746     113.401   \n",
 889 |        "3     109.088     108.214     108.176     107.477     107.332     108.305   \n",
 890 |        "4     110.649     110.649     110.649     110.649     110.649     110.649   \n",
 891 |        "\n",
 892 |        "   add_tp_118  add_tp_119  add_tp_120  add_tp_121  \n",
 893 |        "0     109.161     108.627     108.839     109.081  \n",
 894 |        "1     121.885     119.338     118.589     114.357  \n",
 895 |        "2     116.527     118.449     119.659     120.692  \n",
 896 |        "3     110.405     112.255     114.158      115.68  \n",
 897 |        "4     110.649     110.649     110.649     110.649  \n",
 898 |        "\n",
 899 |        "[5 rows x 1656 columns]"
 900 |       ]
 901 |      },
 902 |      "execution_count": 39,
 903 |      "metadata": {},
 904 |      "output_type": "execute_result"
 905 |     }
 906 |    ],
 907 |    "source": [
 908 |     "comb.head()"
 909 |    ]
 910 |   },
 911 |   {
 912 |    "cell_type": "code",
 913 |    "execution_count": 40,
 914 |    "metadata": {
 915 |     "collapsed": true
 916 |    },
 917 |    "outputs": [],
 918 |    "source": [
 919 |     "def apply_qcut(feat):\n",
 920 |     "    return pd.qcut(comb[feat], 24, labels=False, duplicates='drop')"
 921 |    ]
 922 |   },
 923 |   {
 924 |    "cell_type": "code",
 925 |    "execution_count": 41,
 926 |    "metadata": {
 927 |     "id": "Z1eJlRcDUdrq",
 928 |     "outputId": "4b5cf990-88bb-4147-b6cc-27f8f9af991a"
 929 |    },
 930 |    "outputs": [
 931 |     {
 932 |      "data": {
 933 |       "application/vnd.jupyter.widget-view+json": {
 934 |        "model_id": "3567c0b91c18424a881e581f27602992",
 935 |        "version_major": 2,
 936 |        "version_minor": 0
 937 |       },
 938 |       "text/plain": [
 939 |        "HBox(children=(FloatProgress(value=0.0, max=1653.0), HTML(value='')))"
 940 |       ]
 941 |      },
 942 |      "metadata": {},
 943 |      "output_type": "display_data"
 944 |     },
 945 |     {
 946 |      "name": "stdout",
 947 |      "output_type": "stream",
 948 |      "text": [
 949 |       "\n",
 950 |       "CPU times: user 2.11 s, sys: 1.36 s, total: 3.48 s\n",
 951 |       "Wall time: 18.5 s\n"
 952 |      ]
 953 |     }
 954 |    ],
 955 |    "source": [
 956 |     "%%time\n",
 957 |     "# Binning feaures\n",
 958 |     "other_features = [x for x in comb.columns if x not in ['separator', 'ID', 'location']]\n",
 959 |     "\n",
 960 |     "# Multiprocessing trick: 15 seconds instead of 7 minutes !\n",
 961 |     "binned_data = joblib.Parallel(n_jobs=-1, backend='multiprocessing')(\n",
 962 |     "    joblib.delayed(apply_qcut)(feat) for feat in tqdm_notebook(other_features))\n",
 963 |     "\n",
 964 |     "comb_binned_data = pd.concat(binned_data, axis=1)\n",
 965 |     "comb = pd.concat([comb[['separator', 'ID', 'location']], comb_binned_data], axis=1)\n",
 966 |     "\n",
 967 |     "# Separating train and test from the combined dataframe\n",
 968 |     "train = comb[comb.separator == 0]\n",
 969 |     "test = comb[comb.separator == 1]\n",
 970 |     "train.drop('separator', axis = 1, inplace = True)\n",
 971 |     "test.drop('separator', axis = 1, inplace = True)\n",
 972 |     "\n",
 973 |     "# Creating a list of test ids in the order that they will be trained\n",
 974 |     "testA =  test[test.location == 'A']\n",
 975 |     "testB =  test[test.location == 'B']\n",
 976 |     "testC =  test[test.location == 'C']\n",
 977 |     "testD =  test[test.location == 'D']\n",
 978 |     "testE =  test[test.location == 'E']\n",
 979 |     "\n",
 980 |     "tA, tD, tE, tBC = testA.ID, testD.ID, testE.ID, test[(test.location == 'B') | (test.location == 'C')].ID\n",
 981 |     "test_id = pd.concat([tA, tD, tE, tBC])\n",
 982 |     "\n",
 983 |     "# Adding back target to the train set\n",
 984 |     "train['target'] = target"
 985 |    ]
 986 |   },
 987 |   {
 988 |    "cell_type": "code",
 989 |    "execution_count": 42,
 990 |    "metadata": {
 991 |     "id": "95P-K-ZljXYv",
 992 |     "outputId": "0e8d98eb-fbcb-4100-a4a8-fc434398a399"
 993 |    },
 994 |    "outputs": [
 995 |     {
 996 |      "data": {
 997 |       "application/vnd.jupyter.widget-view+json": {
 998 |        "model_id": "fd94cfe7d8b84087bd429024bb637184",
 999 |        "version_major": 2,
1000 |        "version_minor": 0
1001 |       },
1002 |       "text/plain": [
1003 |        "HBox(children=(FloatProgress(value=0.0, max=25.0), HTML(value='')))"
1004 |       ]
1005 |      },
1006 |      "metadata": {},
1007 |      "output_type": "display_data"
1008 |     },
1009 |     {
1010 |      "name": "stdout",
1011 |      "output_type": "stream",
1012 |      "text": [
1013 |       "\n",
1014 |       "CPU times: user 2h 35min 4s, sys: 5min 2s, total: 2h 40min 7s\n",
1015 |       "Wall time: 42min 26s\n"
1016 |      ]
1017 |     }
1018 |    ],
1019 |    "source": [
1020 |     "%%time\n",
1021 |     "# Creating X and y values\n",
1022 |     "X = train.drop(['ID', 'location', 'target'], axis = 1)\n",
1023 |     "y = target.values\n",
1024 |     "\n",
1025 |     "# Shuffling the X, y values\n",
1026 |     "X, y = shuffle(X, y, random_state = 0)\n",
1027 |     "tes = test.drop(['ID', 'location'], axis = 1)\n",
1028 |     "\n",
1029 |     "# Traing the model across multiple seeds\n",
1030 |     "predictions = []\n",
1031 |     "for i in tqdm_notebook(range(25)):\n",
1032 |     "  cat = CatBoostRegressor(verbose = False, random_seed=i)\n",
1033 |     "  cat.fit(X, y)\n",
1034 |     "\n",
1035 |     "  preds = cat.predict(tes)\n",
1036 |     "  predictions.append(preds)\n",
1037 |     "\n",
1038 |     "# Averaging the predictions\n",
1039 |     "avg_preds = np.mean(predictions, axis = 0)\n",
1040 |     "\n",
1041 |     "# Post processing of the predictions\n",
1042 |     "post_proc = [((((((((((x-0.85)*1.015)-0.85)*1.012)-0.75)*1.0095)-0.55)*1.0065)-0.8)*1.007) for x in avg_preds]\n",
1043 |     "post_proc = predzz = [((x-0.85)*1.015) for x in post_proc]\n",
1044 |     "\n",
1045 |     "# Creating a submission file\n",
1046 |     "sub_df = pd.DataFrame({'ID': test.ID, 'target':post_proc})\n",
1047 |     "sub_df.to_csv('model_2.csv', index = False)"
1048 |    ]
1049 |   },
1050 |   {
1051 |    "cell_type": "code",
1052 |    "execution_count": 43,
1053 |    "metadata": {},
1054 |    "outputs": [
1055 |     {
1056 |      "data": {
1057 |       "text/html": [
1058 |        "<div>\n",
1059 |        "<style scoped>\n",
1060 |        "    .dataframe tbody tr th:only-of-type {\n",
1061 |        "        vertical-align: middle;\n",
1062 |        "    }\n",
1063 |        "\n",
1064 |        "    .dataframe tbody tr th {\n",
1065 |        "        vertical-align: top;\n",
1066 |        "    }\n",
1067 |        "\n",
1068 |        "    .dataframe thead th {\n",
1069 |        "        text-align: right;\n",
1070 |        "    }\n",
1071 |        "</style>\n",
1072 |        "<table border=\"1\" class=\"dataframe\">\n",
1073 |        "  <thead>\n",
1074 |        "    <tr style=\"text-align: right;\">\n",
1075 |        "      <th></th>\n",
1076 |        "      <th>ID</th>\n",
1077 |        "      <th>target</th>\n",
1078 |        "    </tr>\n",
1079 |        "  </thead>\n",
1080 |        "  <tbody>\n",
1081 |        "    <tr>\n",
1082 |        "      <th>15539</th>\n",
1083 |        "      <td>ID_test_0</td>\n",
1084 |        "      <td>154.512605</td>\n",
1085 |        "    </tr>\n",
1086 |        "    <tr>\n",
1087 |        "      <th>15540</th>\n",
1088 |        "      <td>ID_test_1</td>\n",
1089 |        "      <td>117.225585</td>\n",
1090 |        "    </tr>\n",
1091 |        "    <tr>\n",
1092 |        "      <th>15541</th>\n",
1093 |        "      <td>ID_test_10</td>\n",
1094 |        "      <td>26.247779</td>\n",
1095 |        "    </tr>\n",
1096 |        "    <tr>\n",
1097 |        "      <th>15542</th>\n",
1098 |        "      <td>ID_test_100</td>\n",
1099 |        "      <td>63.167372</td>\n",
1100 |        "    </tr>\n",
1101 |        "    <tr>\n",
1102 |        "      <th>15543</th>\n",
1103 |        "      <td>ID_test_1000</td>\n",
1104 |        "      <td>92.044408</td>\n",
1105 |        "    </tr>\n",
1106 |        "    <tr>\n",
1107 |        "      <th>15544</th>\n",
1108 |        "      <td>ID_test_1001</td>\n",
1109 |        "      <td>41.451966</td>\n",
1110 |        "    </tr>\n",
1111 |        "    <tr>\n",
1112 |        "      <th>15545</th>\n",
1113 |        "      <td>ID_test_1002</td>\n",
1114 |        "      <td>84.857269</td>\n",
1115 |        "    </tr>\n",
1116 |        "    <tr>\n",
1117 |        "      <th>15546</th>\n",
1118 |        "      <td>ID_test_1003</td>\n",
1119 |        "      <td>37.807807</td>\n",
1120 |        "    </tr>\n",
1121 |        "    <tr>\n",
1122 |        "      <th>15547</th>\n",
1123 |        "      <td>ID_test_1004</td>\n",
1124 |        "      <td>30.820292</td>\n",
1125 |        "    </tr>\n",
1126 |        "    <tr>\n",
1127 |        "      <th>15548</th>\n",
1128 |        "      <td>ID_test_1005</td>\n",
1129 |        "      <td>45.990773</td>\n",
1130 |        "    </tr>\n",
1131 |        "  </tbody>\n",
1132 |        "</table>\n",
1133 |        "</div>"
1134 |       ],
1135 |       "text/plain": [
1136 |        "                 ID      target\n",
1137 |        "15539     ID_test_0  154.512605\n",
1138 |        "15540     ID_test_1  117.225585\n",
1139 |        "15541    ID_test_10   26.247779\n",
1140 |        "15542   ID_test_100   63.167372\n",
1141 |        "15543  ID_test_1000   92.044408\n",
1142 |        "15544  ID_test_1001   41.451966\n",
1143 |        "15545  ID_test_1002   84.857269\n",
1144 |        "15546  ID_test_1003   37.807807\n",
1145 |        "15547  ID_test_1004   30.820292\n",
1146 |        "15548  ID_test_1005   45.990773"
1147 |       ]
1148 |      },
1149 |      "execution_count": 43,
1150 |      "metadata": {},
1151 |      "output_type": "execute_result"
1152 |     }
1153 |    ],
1154 |    "source": [
1155 |     "sub_df.head(10)"
1156 |    ]
1157 |   },
1158 |   {
1159 |    "cell_type": "code",
1160 |    "execution_count": null,
1161 |    "metadata": {
1162 |     "collapsed": true
1163 |    },
1164 |    "outputs": [],
1165 |    "source": []
1166 |   }
1167 |  ],
1168 |  "metadata": {
1169 |   "kernelspec": {
1170 |    "display_name": "Python 3",
1171 |    "language": "python",
1172 |    "name": "python3"
1173 |   },
1174 |   "language_info": {
1175 |    "codemirror_mode": {
1176 |     "name": "ipython",
1177 |     "version": 3
1178 |    },
1179 |    "file_extension": ".py",
1180 |    "mimetype": "text/x-python",
1181 |    "name": "python",
1182 |    "nbconvert_exporter": "python",
1183 |    "pygments_lexer": "ipython3",
1184 |    "version": "3.6.3"
1185 |   }
1186 |  },
1187 |  "nbformat": 4,
1188 |  "nbformat_minor": 4
1189 | }
1190 | 


--------------------------------------------------------------------------------
/zindi-airqo-cnn-quick.ipynb:
--------------------------------------------------------------------------------
   1 | {
   2 |  "cells": [
   3 |   {
   4 |    "cell_type": "code",
   5 |    "execution_count": 28,
   6 |    "metadata": {
   7 |     "collapsed": true
   8 |    },
   9 |    "outputs": [],
  10 |    "source": [
  11 |     "### CNN output may vary everytime due to reproducibility issues\n",
  12 |     "### Using Kaggle GPU it takes only 22 minutes to run.\n",
  13 |     "### CNN is the most feasible option for implementation as the data size grows.\n",
  14 |     "### It does not require any kind of feature engineering - CNN does auto feature engineering.\n",
  15 |     "### GPU makes crunching a big dataset easier and faster.\n",
  16 |     "### CNN performance improves as the data increases - Also CNN has a good really CV.\n",
  17 |     "\n",
  18 |     "# -> Our CNN can easily handle different features. So even if new meteorological features are added no worries !!!"
  19 |    ]
  20 |   },
  21 |   {
  22 |    "cell_type": "code",
  23 |    "execution_count": 1,
  24 |    "metadata": {
  25 |     "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
  26 |     "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
  27 |     "collapsed": true
  28 |    },
  29 |    "outputs": [],
  30 |    "source": [
  31 |     "import pandas as pd \n",
  32 |     "import numpy as np \n",
  33 |     "from tqdm.notebook import tqdm\n",
  34 |     "import math\n",
  35 |     "import gc\n",
  36 |     "from sklearn.preprocessing import LabelEncoder\n",
  37 |     "import matplotlib.pyplot as plt\n",
  38 |     "from sklearn.model_selection import KFold, train_test_split\n",
  39 |     "\n",
  40 |     "import os\n",
  41 |     "import shutil\n",
  42 |     "import datetime\n",
  43 |     "from tqdm import tqdm, tqdm_notebook\n",
  44 |     "\n",
  45 |     "%matplotlib inline\n",
  46 |     "\n",
  47 |     "pd.set_option(\"display.max_rows\", 200)\n",
  48 |     "pd.set_option(\"display.max_columns\", 200)"
  49 |    ]
  50 |   },
  51 |   {
  52 |    "cell_type": "code",
  53 |    "execution_count": 2,
  54 |    "metadata": {
  55 |     "collapsed": true
  56 |    },
  57 |    "outputs": [],
  58 |    "source": [
  59 |     "from datetime import datetime\n",
  60 |     "\n",
  61 |     "from sklearn.model_selection import train_test_split\n",
  62 |     "from sklearn.metrics import mean_squared_error\n",
  63 |     "from sklearn.preprocessing import StandardScaler\n",
  64 |     "\n",
  65 |     "# import keras\n",
  66 |     "import tensorflow as tf\n",
  67 |     "from tensorflow.keras.models import Sequential, Model\n",
  68 |     "from tensorflow.keras.layers import Dense, Conv1D, Flatten, Dropout, Input, BatchNormalization\n",
  69 |     "from tensorflow.keras.layers import *\n",
  70 |     "from tensorflow.keras import callbacks, optimizers\n",
  71 |     "from tensorflow.keras.optimizers import Adam\n",
  72 |     "import tensorflow.keras.backend as K\n",
  73 |     "from joblib import Parallel, delayed\n",
  74 |     "from functools import partial\n",
  75 |     "import gc\n",
  76 |     "from tensorflow.keras.layers import MaxPooling1D"
  77 |    ]
  78 |   },
  79 |   {
  80 |    "cell_type": "code",
  81 |    "execution_count": 1,
  82 |    "metadata": {
  83 |     "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0",
  84 |     "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a",
  85 |     "collapsed": true
  86 |    },
  87 |    "outputs": [],
  88 |    "source": [
  89 |     "# !mkdir input\n",
  90 |     "# !cp /kaggle/input/zindi-airqo/* ./input"
  91 |    ]
  92 |   },
  93 |   {
  94 |    "cell_type": "code",
  95 |    "execution_count": 4,
  96 |    "metadata": {
  97 |     "collapsed": true
  98 |    },
  99 |    "outputs": [],
 100 |    "source": [
 101 |     "# Loading data\n",
 102 |     "train = pd.read_csv('./input/Train.csv')\n",
 103 |     "test = pd.read_csv('./input/Test.csv')"
 104 |    ]
 105 |   },
 106 |   {
 107 |    "cell_type": "code",
 108 |    "execution_count": 8,
 109 |    "metadata": {},
 110 |    "outputs": [
 111 |     {
 112 |      "name": "stderr",
 113 |      "output_type": "stream",
 114 |      "text": [
 115 |       "/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:5: TqdmDeprecationWarning: This function will be removed in tqdm==5.0.0\n",
 116 |       "Please use `tqdm.notebook.tqdm` instead of `tqdm.tqdm_notebook`\n",
 117 |       "  \"\"\"\n"
 118 |      ]
 119 |     },
 120 |     {
 121 |      "data": {
 122 |       "application/vnd.jupyter.widget-view+json": {
 123 |        "model_id": "6e4983244c3c4439be3e2b17ec9374ab",
 124 |        "version_major": 2,
 125 |        "version_minor": 0
 126 |       },
 127 |       "text/plain": [
 128 |        "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
 129 |       ]
 130 |      },
 131 |      "metadata": {},
 132 |      "output_type": "display_data"
 133 |     },
 134 |     {
 135 |      "name": "stdout",
 136 |      "output_type": "stream",
 137 |      "text": [
 138 |       "\n"
 139 |      ]
 140 |     },
 141 |     {
 142 |      "name": "stderr",
 143 |      "output_type": "stream",
 144 |      "text": [
 145 |       "/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:16: TqdmDeprecationWarning: This function will be removed in tqdm==5.0.0\n",
 146 |       "Please use `tqdm.notebook.tqdm` instead of `tqdm.tqdm_notebook`\n",
 147 |       "  app.launch_new_instance()\n"
 148 |      ]
 149 |     },
 150 |     {
 151 |      "data": {
 152 |       "application/vnd.jupyter.widget-view+json": {
 153 |        "model_id": "266970b55163443f9c9029b908024b24",
 154 |        "version_major": 2,
 155 |        "version_minor": 0
 156 |       },
 157 |       "text/plain": [
 158 |        "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
 159 |       ]
 160 |      },
 161 |      "metadata": {},
 162 |      "output_type": "display_data"
 163 |     },
 164 |     {
 165 |      "name": "stdout",
 166 |      "output_type": "stream",
 167 |      "text": [
 168 |       "\n",
 169 |       "(20574, 857)\n",
 170 |       "144018\n",
 171 |       "CPU times: user 22 s, sys: 2.7 s, total: 24.7 s\n",
 172 |       "Wall time: 33.5 s\n"
 173 |      ]
 174 |     },
 175 |     {
 176 |      "data": {
 177 |       "text/html": [
 178 |        "<div>\n",
 179 |        "<style scoped>\n",
 180 |        "    .dataframe tbody tr th:only-of-type {\n",
 181 |        "        vertical-align: middle;\n",
 182 |        "    }\n",
 183 |        "\n",
 184 |        "    .dataframe tbody tr th {\n",
 185 |        "        vertical-align: top;\n",
 186 |        "    }\n",
 187 |        "\n",
 188 |        "    .dataframe thead th {\n",
 189 |        "        text-align: right;\n",
 190 |        "    }\n",
 191 |        "</style>\n",
 192 |        "<table border=\"1\" class=\"dataframe\">\n",
 193 |        "  <thead>\n",
 194 |        "    <tr style=\"text-align: right;\">\n",
 195 |        "      <th></th>\n",
 196 |        "      <th>ID</th>\n",
 197 |        "      <th>location</th>\n",
 198 |        "      <th>target</th>\n",
 199 |        "      <th>temp</th>\n",
 200 |        "      <th>precip</th>\n",
 201 |        "      <th>rel_humidity</th>\n",
 202 |        "      <th>wind_spd</th>\n",
 203 |        "      <th>atmos_press</th>\n",
 204 |        "      <th>temp_0</th>\n",
 205 |        "      <th>temp_1</th>\n",
 206 |        "      <th>temp_2</th>\n",
 207 |        "      <th>temp_3</th>\n",
 208 |        "      <th>temp_4</th>\n",
 209 |        "      <th>temp_5</th>\n",
 210 |        "      <th>temp_6</th>\n",
 211 |        "      <th>temp_7</th>\n",
 212 |        "      <th>temp_8</th>\n",
 213 |        "      <th>temp_9</th>\n",
 214 |        "      <th>temp_10</th>\n",
 215 |        "      <th>temp_11</th>\n",
 216 |        "      <th>temp_12</th>\n",
 217 |        "      <th>temp_13</th>\n",
 218 |        "      <th>temp_14</th>\n",
 219 |        "      <th>temp_15</th>\n",
 220 |        "      <th>temp_16</th>\n",
 221 |        "      <th>temp_17</th>\n",
 222 |        "      <th>temp_18</th>\n",
 223 |        "      <th>temp_19</th>\n",
 224 |        "      <th>temp_20</th>\n",
 225 |        "      <th>temp_21</th>\n",
 226 |        "      <th>temp_22</th>\n",
 227 |        "      <th>temp_23</th>\n",
 228 |        "      <th>temp_24</th>\n",
 229 |        "      <th>temp_25</th>\n",
 230 |        "      <th>temp_26</th>\n",
 231 |        "      <th>temp_27</th>\n",
 232 |        "      <th>temp_28</th>\n",
 233 |        "      <th>temp_29</th>\n",
 234 |        "      <th>temp_30</th>\n",
 235 |        "      <th>temp_31</th>\n",
 236 |        "      <th>temp_32</th>\n",
 237 |        "      <th>temp_33</th>\n",
 238 |        "      <th>temp_34</th>\n",
 239 |        "      <th>temp_35</th>\n",
 240 |        "      <th>temp_36</th>\n",
 241 |        "      <th>temp_37</th>\n",
 242 |        "      <th>temp_38</th>\n",
 243 |        "      <th>temp_39</th>\n",
 244 |        "      <th>temp_40</th>\n",
 245 |        "      <th>temp_41</th>\n",
 246 |        "      <th>temp_42</th>\n",
 247 |        "      <th>temp_43</th>\n",
 248 |        "      <th>temp_44</th>\n",
 249 |        "      <th>temp_45</th>\n",
 250 |        "      <th>temp_46</th>\n",
 251 |        "      <th>temp_47</th>\n",
 252 |        "      <th>temp_48</th>\n",
 253 |        "      <th>temp_49</th>\n",
 254 |        "      <th>temp_50</th>\n",
 255 |        "      <th>temp_51</th>\n",
 256 |        "      <th>temp_52</th>\n",
 257 |        "      <th>temp_53</th>\n",
 258 |        "      <th>temp_54</th>\n",
 259 |        "      <th>temp_55</th>\n",
 260 |        "      <th>temp_56</th>\n",
 261 |        "      <th>temp_57</th>\n",
 262 |        "      <th>temp_58</th>\n",
 263 |        "      <th>temp_59</th>\n",
 264 |        "      <th>temp_60</th>\n",
 265 |        "      <th>temp_61</th>\n",
 266 |        "      <th>temp_62</th>\n",
 267 |        "      <th>temp_63</th>\n",
 268 |        "      <th>temp_64</th>\n",
 269 |        "      <th>temp_65</th>\n",
 270 |        "      <th>temp_66</th>\n",
 271 |        "      <th>temp_67</th>\n",
 272 |        "      <th>temp_68</th>\n",
 273 |        "      <th>temp_69</th>\n",
 274 |        "      <th>temp_70</th>\n",
 275 |        "      <th>temp_71</th>\n",
 276 |        "      <th>temp_72</th>\n",
 277 |        "      <th>temp_73</th>\n",
 278 |        "      <th>temp_74</th>\n",
 279 |        "      <th>temp_75</th>\n",
 280 |        "      <th>temp_76</th>\n",
 281 |        "      <th>temp_77</th>\n",
 282 |        "      <th>temp_78</th>\n",
 283 |        "      <th>temp_79</th>\n",
 284 |        "      <th>temp_80</th>\n",
 285 |        "      <th>temp_81</th>\n",
 286 |        "      <th>temp_82</th>\n",
 287 |        "      <th>temp_83</th>\n",
 288 |        "      <th>temp_84</th>\n",
 289 |        "      <th>temp_85</th>\n",
 290 |        "      <th>temp_86</th>\n",
 291 |        "      <th>temp_87</th>\n",
 292 |        "      <th>temp_88</th>\n",
 293 |        "      <th>temp_89</th>\n",
 294 |        "      <th>temp_90</th>\n",
 295 |        "      <th>temp_91</th>\n",
 296 |        "      <th>...</th>\n",
 297 |        "      <th>wind_dir_sin_21</th>\n",
 298 |        "      <th>wind_dir_sin_22</th>\n",
 299 |        "      <th>wind_dir_sin_23</th>\n",
 300 |        "      <th>wind_dir_sin_24</th>\n",
 301 |        "      <th>wind_dir_sin_25</th>\n",
 302 |        "      <th>wind_dir_sin_26</th>\n",
 303 |        "      <th>wind_dir_sin_27</th>\n",
 304 |        "      <th>wind_dir_sin_28</th>\n",
 305 |        "      <th>wind_dir_sin_29</th>\n",
 306 |        "      <th>wind_dir_sin_30</th>\n",
 307 |        "      <th>wind_dir_sin_31</th>\n",
 308 |        "      <th>wind_dir_sin_32</th>\n",
 309 |        "      <th>wind_dir_sin_33</th>\n",
 310 |        "      <th>wind_dir_sin_34</th>\n",
 311 |        "      <th>wind_dir_sin_35</th>\n",
 312 |        "      <th>wind_dir_sin_36</th>\n",
 313 |        "      <th>wind_dir_sin_37</th>\n",
 314 |        "      <th>wind_dir_sin_38</th>\n",
 315 |        "      <th>wind_dir_sin_39</th>\n",
 316 |        "      <th>wind_dir_sin_40</th>\n",
 317 |        "      <th>wind_dir_sin_41</th>\n",
 318 |        "      <th>wind_dir_sin_42</th>\n",
 319 |        "      <th>wind_dir_sin_43</th>\n",
 320 |        "      <th>wind_dir_sin_44</th>\n",
 321 |        "      <th>wind_dir_sin_45</th>\n",
 322 |        "      <th>wind_dir_sin_46</th>\n",
 323 |        "      <th>wind_dir_sin_47</th>\n",
 324 |        "      <th>wind_dir_sin_48</th>\n",
 325 |        "      <th>wind_dir_sin_49</th>\n",
 326 |        "      <th>wind_dir_sin_50</th>\n",
 327 |        "      <th>wind_dir_sin_51</th>\n",
 328 |        "      <th>wind_dir_sin_52</th>\n",
 329 |        "      <th>wind_dir_sin_53</th>\n",
 330 |        "      <th>wind_dir_sin_54</th>\n",
 331 |        "      <th>wind_dir_sin_55</th>\n",
 332 |        "      <th>wind_dir_sin_56</th>\n",
 333 |        "      <th>wind_dir_sin_57</th>\n",
 334 |        "      <th>wind_dir_sin_58</th>\n",
 335 |        "      <th>wind_dir_sin_59</th>\n",
 336 |        "      <th>wind_dir_sin_60</th>\n",
 337 |        "      <th>wind_dir_sin_61</th>\n",
 338 |        "      <th>wind_dir_sin_62</th>\n",
 339 |        "      <th>wind_dir_sin_63</th>\n",
 340 |        "      <th>wind_dir_sin_64</th>\n",
 341 |        "      <th>wind_dir_sin_65</th>\n",
 342 |        "      <th>wind_dir_sin_66</th>\n",
 343 |        "      <th>wind_dir_sin_67</th>\n",
 344 |        "      <th>wind_dir_sin_68</th>\n",
 345 |        "      <th>wind_dir_sin_69</th>\n",
 346 |        "      <th>wind_dir_sin_70</th>\n",
 347 |        "      <th>wind_dir_sin_71</th>\n",
 348 |        "      <th>wind_dir_sin_72</th>\n",
 349 |        "      <th>wind_dir_sin_73</th>\n",
 350 |        "      <th>wind_dir_sin_74</th>\n",
 351 |        "      <th>wind_dir_sin_75</th>\n",
 352 |        "      <th>wind_dir_sin_76</th>\n",
 353 |        "      <th>wind_dir_sin_77</th>\n",
 354 |        "      <th>wind_dir_sin_78</th>\n",
 355 |        "      <th>wind_dir_sin_79</th>\n",
 356 |        "      <th>wind_dir_sin_80</th>\n",
 357 |        "      <th>wind_dir_sin_81</th>\n",
 358 |        "      <th>wind_dir_sin_82</th>\n",
 359 |        "      <th>wind_dir_sin_83</th>\n",
 360 |        "      <th>wind_dir_sin_84</th>\n",
 361 |        "      <th>wind_dir_sin_85</th>\n",
 362 |        "      <th>wind_dir_sin_86</th>\n",
 363 |        "      <th>wind_dir_sin_87</th>\n",
 364 |        "      <th>wind_dir_sin_88</th>\n",
 365 |        "      <th>wind_dir_sin_89</th>\n",
 366 |        "      <th>wind_dir_sin_90</th>\n",
 367 |        "      <th>wind_dir_sin_91</th>\n",
 368 |        "      <th>wind_dir_sin_92</th>\n",
 369 |        "      <th>wind_dir_sin_93</th>\n",
 370 |        "      <th>wind_dir_sin_94</th>\n",
 371 |        "      <th>wind_dir_sin_95</th>\n",
 372 |        "      <th>wind_dir_sin_96</th>\n",
 373 |        "      <th>wind_dir_sin_97</th>\n",
 374 |        "      <th>wind_dir_sin_98</th>\n",
 375 |        "      <th>wind_dir_sin_99</th>\n",
 376 |        "      <th>wind_dir_sin_100</th>\n",
 377 |        "      <th>wind_dir_sin_101</th>\n",
 378 |        "      <th>wind_dir_sin_102</th>\n",
 379 |        "      <th>wind_dir_sin_103</th>\n",
 380 |        "      <th>wind_dir_sin_104</th>\n",
 381 |        "      <th>wind_dir_sin_105</th>\n",
 382 |        "      <th>wind_dir_sin_106</th>\n",
 383 |        "      <th>wind_dir_sin_107</th>\n",
 384 |        "      <th>wind_dir_sin_108</th>\n",
 385 |        "      <th>wind_dir_sin_109</th>\n",
 386 |        "      <th>wind_dir_sin_110</th>\n",
 387 |        "      <th>wind_dir_sin_111</th>\n",
 388 |        "      <th>wind_dir_sin_112</th>\n",
 389 |        "      <th>wind_dir_sin_113</th>\n",
 390 |        "      <th>wind_dir_sin_114</th>\n",
 391 |        "      <th>wind_dir_sin_115</th>\n",
 392 |        "      <th>wind_dir_sin_116</th>\n",
 393 |        "      <th>wind_dir_sin_117</th>\n",
 394 |        "      <th>wind_dir_sin_118</th>\n",
 395 |        "      <th>wind_dir_sin_119</th>\n",
 396 |        "      <th>wind_dir_sin_120</th>\n",
 397 |        "    </tr>\n",
 398 |        "  </thead>\n",
 399 |        "  <tbody>\n",
 400 |        "    <tr>\n",
 401 |        "      <th>0</th>\n",
 402 |        "      <td>ID_train_0</td>\n",
 403 |        "      <td>C</td>\n",
 404 |        "      <td>45.126304</td>\n",
 405 |        "      <td>NaN</td>\n",
 406 |        "      <td>NaN</td>\n",
 407 |        "      <td>NaN</td>\n",
 408 |        "      <td>NaN</td>\n",
 409 |        "      <td>NaN</td>\n",
 410 |        "      <td>22.533333</td>\n",
 411 |        "      <td>21.716667</td>\n",
 412 |        "      <td>20.833333</td>\n",
 413 |        "      <td>20.983333</td>\n",
 414 |        "      <td>20.875000</td>\n",
 415 |        "      <td>20.141667</td>\n",
 416 |        "      <td>19.375000</td>\n",
 417 |        "      <td>18.791667</td>\n",
 418 |        "      <td>18.775000</td>\n",
 419 |        "      <td>18.641667</td>\n",
 420 |        "      <td>18.558333</td>\n",
 421 |        "      <td>18.533333</td>\n",
 422 |        "      <td>19.608333</td>\n",
 423 |        "      <td>21.916667</td>\n",
 424 |        "      <td>24.716667</td>\n",
 425 |        "      <td>26.658333</td>\n",
 426 |        "      <td>29.175000</td>\n",
 427 |        "      <td>30.700000</td>\n",
 428 |        "      <td>31.433333</td>\n",
 429 |        "      <td>32.333333</td>\n",
 430 |        "      <td>33.008333</td>\n",
 431 |        "      <td>33.391667</td>\n",
 432 |        "      <td>33.616667</td>\n",
 433 |        "      <td>31.091667</td>\n",
 434 |        "      <td>27.550000</td>\n",
 435 |        "      <td>26.658333</td>\n",
 436 |        "      <td>25.675000</td>\n",
 437 |        "      <td>26.425000</td>\n",
 438 |        "      <td>23.783333</td>\n",
 439 |        "      <td>22.416667</td>\n",
 440 |        "      <td>22.041667</td>\n",
 441 |        "      <td>21.008333</td>\n",
 442 |        "      <td>21.475000</td>\n",
 443 |        "      <td>21.825000</td>\n",
 444 |        "      <td>21.158333</td>\n",
 445 |        "      <td>22.308333</td>\n",
 446 |        "      <td>23.116667</td>\n",
 447 |        "      <td>22.333333</td>\n",
 448 |        "      <td>23.850000</td>\n",
 449 |        "      <td>23.825000</td>\n",
 450 |        "      <td>24.983333</td>\n",
 451 |        "      <td>28.050000</td>\n",
 452 |        "      <td>30.841667</td>\n",
 453 |        "      <td>31.991667</td>\n",
 454 |        "      <td>32.041667</td>\n",
 455 |        "      <td>32.000000</td>\n",
 456 |        "      <td>31.666667</td>\n",
 457 |        "      <td>29.158333</td>\n",
 458 |        "      <td>25.866667</td>\n",
 459 |        "      <td>24.300000</td>\n",
 460 |        "      <td>23.683333</td>\n",
 461 |        "      <td>23.758333</td>\n",
 462 |        "      <td>22.991667</td>\n",
 463 |        "      <td>21.891667</td>\n",
 464 |        "      <td>21.158333</td>\n",
 465 |        "      <td>20.416667</td>\n",
 466 |        "      <td>19.533333</td>\n",
 467 |        "      <td>19.125000</td>\n",
 468 |        "      <td>18.766667</td>\n",
 469 |        "      <td>17.983333</td>\n",
 470 |        "      <td>18.458333</td>\n",
 471 |        "      <td>21.933333</td>\n",
 472 |        "      <td>23.791667</td>\n",
 473 |        "      <td>25.558333</td>\n",
 474 |        "      <td>27.758333</td>\n",
 475 |        "      <td>29.658333</td>\n",
 476 |        "      <td>31.391667</td>\n",
 477 |        "      <td>31.883333</td>\n",
 478 |        "      <td>32.358333</td>\n",
 479 |        "      <td>32.708333</td>\n",
 480 |        "      <td>31.983333</td>\n",
 481 |        "      <td>30.850000</td>\n",
 482 |        "      <td>28.800</td>\n",
 483 |        "      <td>26.491667</td>\n",
 484 |        "      <td>24.866667</td>\n",
 485 |        "      <td>24.366667</td>\n",
 486 |        "      <td>23.025000</td>\n",
 487 |        "      <td>22.325000</td>\n",
 488 |        "      <td>21.650000</td>\n",
 489 |        "      <td>20.750000</td>\n",
 490 |        "      <td>20.475000</td>\n",
 491 |        "      <td>19.641667</td>\n",
 492 |        "      <td>19.516667</td>\n",
 493 |        "      <td>19.575000</td>\n",
 494 |        "      <td>20.000000</td>\n",
 495 |        "      <td>23.358333</td>\n",
 496 |        "      <td>25.608333</td>\n",
 497 |        "      <td>26.883333</td>\n",
 498 |        "      <td>26.358333</td>\n",
 499 |        "      <td>27.325000</td>\n",
 500 |        "      <td>29.008333</td>\n",
 501 |        "      <td>28.433333</td>\n",
 502 |        "      <td>...</td>\n",
 503 |        "      <td>-0.141620</td>\n",
 504 |        "      <td>0.677692</td>\n",
 505 |        "      <td>-0.691042</td>\n",
 506 |        "      <td>-0.570434</td>\n",
 507 |        "      <td>-0.621553</td>\n",
 508 |        "      <td>0.054916</td>\n",
 509 |        "      <td>-0.068411</td>\n",
 510 |        "      <td>-0.427471</td>\n",
 511 |        "      <td>-0.748119</td>\n",
 512 |        "      <td>0.795807</td>\n",
 513 |        "      <td>0.997073</td>\n",
 514 |        "      <td>-0.105681</td>\n",
 515 |        "      <td>-0.170296</td>\n",
 516 |        "      <td>0.850561</td>\n",
 517 |        "      <td>0.971100</td>\n",
 518 |        "      <td>-0.002620</td>\n",
 519 |        "      <td>-0.897881</td>\n",
 520 |        "      <td>0.855572</td>\n",
 521 |        "      <td>0.562276</td>\n",
 522 |        "      <td>0.606049</td>\n",
 523 |        "      <td>0.902810</td>\n",
 524 |        "      <td>-0.853162</td>\n",
 525 |        "      <td>-0.186615</td>\n",
 526 |        "      <td>-0.032798</td>\n",
 527 |        "      <td>0.937824</td>\n",
 528 |        "      <td>-0.317277</td>\n",
 529 |        "      <td>-0.451505</td>\n",
 530 |        "      <td>-0.797416</td>\n",
 531 |        "      <td>-0.159037</td>\n",
 532 |        "      <td>-0.842704</td>\n",
 533 |        "      <td>0.590000</td>\n",
 534 |        "      <td>0.249485</td>\n",
 535 |        "      <td>0.643785</td>\n",
 536 |        "      <td>0.369674</td>\n",
 537 |        "      <td>-0.910274</td>\n",
 538 |        "      <td>-0.390897</td>\n",
 539 |        "      <td>0.531411</td>\n",
 540 |        "      <td>-0.117671</td>\n",
 541 |        "      <td>0.682072</td>\n",
 542 |        "      <td>0.997411</td>\n",
 543 |        "      <td>-0.445040</td>\n",
 544 |        "      <td>0.641554</td>\n",
 545 |        "      <td>0.917998</td>\n",
 546 |        "      <td>0.647900</td>\n",
 547 |        "      <td>0.843400</td>\n",
 548 |        "      <td>0.831011</td>\n",
 549 |        "      <td>-0.780758</td>\n",
 550 |        "      <td>-0.402907</td>\n",
 551 |        "      <td>-0.970055</td>\n",
 552 |        "      <td>-0.775132</td>\n",
 553 |        "      <td>-0.737531</td>\n",
 554 |        "      <td>-0.877086</td>\n",
 555 |        "      <td>0.706791</td>\n",
 556 |        "      <td>-0.991253</td>\n",
 557 |        "      <td>0.870372</td>\n",
 558 |        "      <td>0.235040</td>\n",
 559 |        "      <td>-0.399780</td>\n",
 560 |        "      <td>-0.890181</td>\n",
 561 |        "      <td>-0.429068</td>\n",
 562 |        "      <td>0.536505</td>\n",
 563 |        "      <td>0.849812</td>\n",
 564 |        "      <td>-0.921784</td>\n",
 565 |        "      <td>0.890076</td>\n",
 566 |        "      <td>0.325722</td>\n",
 567 |        "      <td>-0.968946</td>\n",
 568 |        "      <td>0.422955</td>\n",
 569 |        "      <td>-0.908680</td>\n",
 570 |        "      <td>-0.238531</td>\n",
 571 |        "      <td>0.994556</td>\n",
 572 |        "      <td>-0.784712</td>\n",
 573 |        "      <td>-0.327369</td>\n",
 574 |        "      <td>0.793215</td>\n",
 575 |        "      <td>0.921165</td>\n",
 576 |        "      <td>0.900036</td>\n",
 577 |        "      <td>-0.244630</td>\n",
 578 |        "      <td>-0.858170</td>\n",
 579 |        "      <td>-0.980876</td>\n",
 580 |        "      <td>0.890335</td>\n",
 581 |        "      <td>-0.992381</td>\n",
 582 |        "      <td>-0.861434</td>\n",
 583 |        "      <td>-0.629305</td>\n",
 584 |        "      <td>-0.290513</td>\n",
 585 |        "      <td>0.999024</td>\n",
 586 |        "      <td>-0.985627</td>\n",
 587 |        "      <td>0.403930</td>\n",
 588 |        "      <td>-0.778758</td>\n",
 589 |        "      <td>-0.623189</td>\n",
 590 |        "      <td>0.921356</td>\n",
 591 |        "      <td>-0.949976</td>\n",
 592 |        "      <td>-0.938622</td>\n",
 593 |        "      <td>0.539989</td>\n",
 594 |        "      <td>0.833571</td>\n",
 595 |        "      <td>-0.566473</td>\n",
 596 |        "      <td>0.249451</td>\n",
 597 |        "      <td>0.703657</td>\n",
 598 |        "      <td>-0.775581</td>\n",
 599 |        "      <td>0.150566</td>\n",
 600 |        "      <td>0.117369</td>\n",
 601 |        "      <td>0.635717</td>\n",
 602 |        "      <td>-0.947955</td>\n",
 603 |        "    </tr>\n",
 604 |        "    <tr>\n",
 605 |        "      <th>1</th>\n",
 606 |        "      <td>ID_train_1</td>\n",
 607 |        "      <td>D</td>\n",
 608 |        "      <td>79.131702</td>\n",
 609 |        "      <td>NaN</td>\n",
 610 |        "      <td>NaN</td>\n",
 611 |        "      <td>NaN</td>\n",
 612 |        "      <td>NaN</td>\n",
 613 |        "      <td>NaN</td>\n",
 614 |        "      <td>22.533333</td>\n",
 615 |        "      <td>21.716667</td>\n",
 616 |        "      <td>20.833333</td>\n",
 617 |        "      <td>20.983333</td>\n",
 618 |        "      <td>20.875000</td>\n",
 619 |        "      <td>20.141667</td>\n",
 620 |        "      <td>19.375000</td>\n",
 621 |        "      <td>18.791667</td>\n",
 622 |        "      <td>18.775000</td>\n",
 623 |        "      <td>18.641667</td>\n",
 624 |        "      <td>18.558333</td>\n",
 625 |        "      <td>18.533333</td>\n",
 626 |        "      <td>19.608333</td>\n",
 627 |        "      <td>21.916667</td>\n",
 628 |        "      <td>24.716667</td>\n",
 629 |        "      <td>26.658333</td>\n",
 630 |        "      <td>29.175000</td>\n",
 631 |        "      <td>30.700000</td>\n",
 632 |        "      <td>31.433333</td>\n",
 633 |        "      <td>32.333333</td>\n",
 634 |        "      <td>33.008333</td>\n",
 635 |        "      <td>33.391667</td>\n",
 636 |        "      <td>33.616667</td>\n",
 637 |        "      <td>31.091667</td>\n",
 638 |        "      <td>27.550000</td>\n",
 639 |        "      <td>26.658333</td>\n",
 640 |        "      <td>25.675000</td>\n",
 641 |        "      <td>26.425000</td>\n",
 642 |        "      <td>23.783333</td>\n",
 643 |        "      <td>22.416667</td>\n",
 644 |        "      <td>22.041667</td>\n",
 645 |        "      <td>21.008333</td>\n",
 646 |        "      <td>21.475000</td>\n",
 647 |        "      <td>21.825000</td>\n",
 648 |        "      <td>21.158333</td>\n",
 649 |        "      <td>22.308333</td>\n",
 650 |        "      <td>23.116667</td>\n",
 651 |        "      <td>22.333333</td>\n",
 652 |        "      <td>23.850000</td>\n",
 653 |        "      <td>23.825000</td>\n",
 654 |        "      <td>24.983333</td>\n",
 655 |        "      <td>28.050000</td>\n",
 656 |        "      <td>30.841667</td>\n",
 657 |        "      <td>31.991667</td>\n",
 658 |        "      <td>32.041667</td>\n",
 659 |        "      <td>32.000000</td>\n",
 660 |        "      <td>31.666667</td>\n",
 661 |        "      <td>29.158333</td>\n",
 662 |        "      <td>25.866667</td>\n",
 663 |        "      <td>24.300000</td>\n",
 664 |        "      <td>23.683333</td>\n",
 665 |        "      <td>23.758333</td>\n",
 666 |        "      <td>22.991667</td>\n",
 667 |        "      <td>21.891667</td>\n",
 668 |        "      <td>21.158333</td>\n",
 669 |        "      <td>20.416667</td>\n",
 670 |        "      <td>19.533333</td>\n",
 671 |        "      <td>19.125000</td>\n",
 672 |        "      <td>18.766667</td>\n",
 673 |        "      <td>17.983333</td>\n",
 674 |        "      <td>18.458333</td>\n",
 675 |        "      <td>21.933333</td>\n",
 676 |        "      <td>23.791667</td>\n",
 677 |        "      <td>25.558333</td>\n",
 678 |        "      <td>27.758333</td>\n",
 679 |        "      <td>29.658333</td>\n",
 680 |        "      <td>31.391667</td>\n",
 681 |        "      <td>31.883333</td>\n",
 682 |        "      <td>32.358333</td>\n",
 683 |        "      <td>32.708333</td>\n",
 684 |        "      <td>31.983333</td>\n",
 685 |        "      <td>30.850000</td>\n",
 686 |        "      <td>28.800</td>\n",
 687 |        "      <td>26.491667</td>\n",
 688 |        "      <td>24.866667</td>\n",
 689 |        "      <td>24.366667</td>\n",
 690 |        "      <td>23.025000</td>\n",
 691 |        "      <td>22.325000</td>\n",
 692 |        "      <td>21.650000</td>\n",
 693 |        "      <td>20.750000</td>\n",
 694 |        "      <td>20.475000</td>\n",
 695 |        "      <td>19.641667</td>\n",
 696 |        "      <td>19.516667</td>\n",
 697 |        "      <td>19.575000</td>\n",
 698 |        "      <td>20.000000</td>\n",
 699 |        "      <td>23.358333</td>\n",
 700 |        "      <td>25.608333</td>\n",
 701 |        "      <td>26.883333</td>\n",
 702 |        "      <td>26.358333</td>\n",
 703 |        "      <td>27.325000</td>\n",
 704 |        "      <td>29.008333</td>\n",
 705 |        "      <td>28.433333</td>\n",
 706 |        "      <td>...</td>\n",
 707 |        "      <td>-0.141620</td>\n",
 708 |        "      <td>0.677692</td>\n",
 709 |        "      <td>-0.691042</td>\n",
 710 |        "      <td>-0.570434</td>\n",
 711 |        "      <td>-0.621553</td>\n",
 712 |        "      <td>0.054916</td>\n",
 713 |        "      <td>-0.068411</td>\n",
 714 |        "      <td>-0.427471</td>\n",
 715 |        "      <td>-0.748119</td>\n",
 716 |        "      <td>0.795807</td>\n",
 717 |        "      <td>0.997073</td>\n",
 718 |        "      <td>-0.105681</td>\n",
 719 |        "      <td>-0.170296</td>\n",
 720 |        "      <td>0.850561</td>\n",
 721 |        "      <td>0.971100</td>\n",
 722 |        "      <td>-0.002620</td>\n",
 723 |        "      <td>-0.897881</td>\n",
 724 |        "      <td>0.855572</td>\n",
 725 |        "      <td>0.562276</td>\n",
 726 |        "      <td>0.606049</td>\n",
 727 |        "      <td>0.902810</td>\n",
 728 |        "      <td>-0.853162</td>\n",
 729 |        "      <td>-0.186615</td>\n",
 730 |        "      <td>-0.032798</td>\n",
 731 |        "      <td>0.937824</td>\n",
 732 |        "      <td>-0.317277</td>\n",
 733 |        "      <td>-0.451505</td>\n",
 734 |        "      <td>-0.797416</td>\n",
 735 |        "      <td>-0.159037</td>\n",
 736 |        "      <td>-0.842704</td>\n",
 737 |        "      <td>0.590000</td>\n",
 738 |        "      <td>0.249485</td>\n",
 739 |        "      <td>0.643785</td>\n",
 740 |        "      <td>0.369674</td>\n",
 741 |        "      <td>-0.910274</td>\n",
 742 |        "      <td>-0.390897</td>\n",
 743 |        "      <td>0.531411</td>\n",
 744 |        "      <td>-0.117671</td>\n",
 745 |        "      <td>0.682072</td>\n",
 746 |        "      <td>0.997411</td>\n",
 747 |        "      <td>-0.445040</td>\n",
 748 |        "      <td>0.641554</td>\n",
 749 |        "      <td>0.917998</td>\n",
 750 |        "      <td>0.647900</td>\n",
 751 |        "      <td>0.843400</td>\n",
 752 |        "      <td>0.831011</td>\n",
 753 |        "      <td>-0.780758</td>\n",
 754 |        "      <td>-0.402907</td>\n",
 755 |        "      <td>-0.970055</td>\n",
 756 |        "      <td>-0.775132</td>\n",
 757 |        "      <td>-0.737531</td>\n",
 758 |        "      <td>-0.877086</td>\n",
 759 |        "      <td>0.706791</td>\n",
 760 |        "      <td>-0.991253</td>\n",
 761 |        "      <td>0.870372</td>\n",
 762 |        "      <td>0.235040</td>\n",
 763 |        "      <td>-0.399780</td>\n",
 764 |        "      <td>-0.890181</td>\n",
 765 |        "      <td>-0.429068</td>\n",
 766 |        "      <td>0.536505</td>\n",
 767 |        "      <td>0.849812</td>\n",
 768 |        "      <td>-0.921784</td>\n",
 769 |        "      <td>0.890076</td>\n",
 770 |        "      <td>0.325722</td>\n",
 771 |        "      <td>-0.968946</td>\n",
 772 |        "      <td>0.422955</td>\n",
 773 |        "      <td>-0.908680</td>\n",
 774 |        "      <td>-0.238531</td>\n",
 775 |        "      <td>0.994556</td>\n",
 776 |        "      <td>-0.784712</td>\n",
 777 |        "      <td>-0.327369</td>\n",
 778 |        "      <td>0.793215</td>\n",
 779 |        "      <td>0.921165</td>\n",
 780 |        "      <td>0.900036</td>\n",
 781 |        "      <td>-0.244630</td>\n",
 782 |        "      <td>-0.858170</td>\n",
 783 |        "      <td>-0.980876</td>\n",
 784 |        "      <td>0.890335</td>\n",
 785 |        "      <td>-0.992381</td>\n",
 786 |        "      <td>-0.861434</td>\n",
 787 |        "      <td>-0.629305</td>\n",
 788 |        "      <td>-0.290513</td>\n",
 789 |        "      <td>0.999024</td>\n",
 790 |        "      <td>-0.985627</td>\n",
 791 |        "      <td>0.923642</td>\n",
 792 |        "      <td>-0.553773</td>\n",
 793 |        "      <td>0.455011</td>\n",
 794 |        "      <td>-0.989001</td>\n",
 795 |        "      <td>0.211568</td>\n",
 796 |        "      <td>-0.230857</td>\n",
 797 |        "      <td>0.971538</td>\n",
 798 |        "      <td>0.669779</td>\n",
 799 |        "      <td>-0.839579</td>\n",
 800 |        "      <td>-0.212372</td>\n",
 801 |        "      <td>-0.029492</td>\n",
 802 |        "      <td>0.903239</td>\n",
 803 |        "      <td>0.881671</td>\n",
 804 |        "      <td>0.304360</td>\n",
 805 |        "      <td>-0.955722</td>\n",
 806 |        "      <td>0.996240</td>\n",
 807 |        "    </tr>\n",
 808 |        "    <tr>\n",
 809 |        "      <th>2</th>\n",
 810 |        "      <td>ID_train_10</td>\n",
 811 |        "      <td>A</td>\n",
 812 |        "      <td>32.661304</td>\n",
 813 |        "      <td>NaN</td>\n",
 814 |        "      <td>NaN</td>\n",
 815 |        "      <td>NaN</td>\n",
 816 |        "      <td>NaN</td>\n",
 817 |        "      <td>NaN</td>\n",
 818 |        "      <td>28.975000</td>\n",
 819 |        "      <td>27.950000</td>\n",
 820 |        "      <td>29.600000</td>\n",
 821 |        "      <td>26.425000</td>\n",
 822 |        "      <td>22.091667</td>\n",
 823 |        "      <td>21.775000</td>\n",
 824 |        "      <td>22.333333</td>\n",
 825 |        "      <td>21.950000</td>\n",
 826 |        "      <td>20.441667</td>\n",
 827 |        "      <td>20.441667</td>\n",
 828 |        "      <td>20.950000</td>\n",
 829 |        "      <td>19.800000</td>\n",
 830 |        "      <td>19.591667</td>\n",
 831 |        "      <td>19.575000</td>\n",
 832 |        "      <td>19.516667</td>\n",
 833 |        "      <td>19.550000</td>\n",
 834 |        "      <td>19.783333</td>\n",
 835 |        "      <td>19.908333</td>\n",
 836 |        "      <td>19.516667</td>\n",
 837 |        "      <td>19.666667</td>\n",
 838 |        "      <td>20.683333</td>\n",
 839 |        "      <td>22.491667</td>\n",
 840 |        "      <td>23.708333</td>\n",
 841 |        "      <td>24.900000</td>\n",
 842 |        "      <td>26.058333</td>\n",
 843 |        "      <td>27.325000</td>\n",
 844 |        "      <td>27.866667</td>\n",
 845 |        "      <td>28.291667</td>\n",
 846 |        "      <td>22.136364</td>\n",
 847 |        "      <td>19.141667</td>\n",
 848 |        "      <td>19.991667</td>\n",
 849 |        "      <td>20.216667</td>\n",
 850 |        "      <td>20.433333</td>\n",
 851 |        "      <td>20.466667</td>\n",
 852 |        "      <td>20.800000</td>\n",
 853 |        "      <td>20.958333</td>\n",
 854 |        "      <td>20.641667</td>\n",
 855 |        "      <td>20.033333</td>\n",
 856 |        "      <td>19.825000</td>\n",
 857 |        "      <td>19.616667</td>\n",
 858 |        "      <td>19.225000</td>\n",
 859 |        "      <td>18.716667</td>\n",
 860 |        "      <td>18.458333</td>\n",
 861 |        "      <td>19.000000</td>\n",
 862 |        "      <td>20.758333</td>\n",
 863 |        "      <td>27.358333</td>\n",
 864 |        "      <td>24.491667</td>\n",
 865 |        "      <td>26.350000</td>\n",
 866 |        "      <td>27.950000</td>\n",
 867 |        "      <td>29.166667</td>\n",
 868 |        "      <td>30.066667</td>\n",
 869 |        "      <td>30.550000</td>\n",
 870 |        "      <td>30.400000</td>\n",
 871 |        "      <td>30.008333</td>\n",
 872 |        "      <td>28.908333</td>\n",
 873 |        "      <td>26.700000</td>\n",
 874 |        "      <td>25.716667</td>\n",
 875 |        "      <td>24.858333</td>\n",
 876 |        "      <td>24.291667</td>\n",
 877 |        "      <td>23.308333</td>\n",
 878 |        "      <td>23.008333</td>\n",
 879 |        "      <td>22.041667</td>\n",
 880 |        "      <td>21.375000</td>\n",
 881 |        "      <td>21.133333</td>\n",
 882 |        "      <td>21.158333</td>\n",
 883 |        "      <td>20.658333</td>\n",
 884 |        "      <td>20.766667</td>\n",
 885 |        "      <td>21.633333</td>\n",
 886 |        "      <td>22.708333</td>\n",
 887 |        "      <td>24.808333</td>\n",
 888 |        "      <td>27.108333</td>\n",
 889 |        "      <td>28.775000</td>\n",
 890 |        "      <td>29.475</td>\n",
 891 |        "      <td>29.766667</td>\n",
 892 |        "      <td>29.875000</td>\n",
 893 |        "      <td>29.150000</td>\n",
 894 |        "      <td>27.716667</td>\n",
 895 |        "      <td>27.491667</td>\n",
 896 |        "      <td>26.483333</td>\n",
 897 |        "      <td>25.475000</td>\n",
 898 |        "      <td>25.008333</td>\n",
 899 |        "      <td>24.600000</td>\n",
 900 |        "      <td>24.033333</td>\n",
 901 |        "      <td>23.358333</td>\n",
 902 |        "      <td>22.366667</td>\n",
 903 |        "      <td>22.608333</td>\n",
 904 |        "      <td>22.741667</td>\n",
 905 |        "      <td>21.908333</td>\n",
 906 |        "      <td>21.550000</td>\n",
 907 |        "      <td>21.758333</td>\n",
 908 |        "      <td>20.316667</td>\n",
 909 |        "      <td>20.650000</td>\n",
 910 |        "      <td>...</td>\n",
 911 |        "      <td>0.992270</td>\n",
 912 |        "      <td>-0.744836</td>\n",
 913 |        "      <td>0.312328</td>\n",
 914 |        "      <td>-0.613390</td>\n",
 915 |        "      <td>-0.691339</td>\n",
 916 |        "      <td>-0.996273</td>\n",
 917 |        "      <td>0.877801</td>\n",
 918 |        "      <td>-0.619629</td>\n",
 919 |        "      <td>-0.852653</td>\n",
 920 |        "      <td>0.952294</td>\n",
 921 |        "      <td>0.272820</td>\n",
 922 |        "      <td>0.887248</td>\n",
 923 |        "      <td>0.161753</td>\n",
 924 |        "      <td>0.710564</td>\n",
 925 |        "      <td>0.493320</td>\n",
 926 |        "      <td>0.831038</td>\n",
 927 |        "      <td>0.922995</td>\n",
 928 |        "      <td>-0.206542</td>\n",
 929 |        "      <td>-0.905140</td>\n",
 930 |        "      <td>0.261427</td>\n",
 931 |        "      <td>-0.834561</td>\n",
 932 |        "      <td>0.999585</td>\n",
 933 |        "      <td>0.969601</td>\n",
 934 |        "      <td>-0.304580</td>\n",
 935 |        "      <td>-0.134351</td>\n",
 936 |        "      <td>0.926241</td>\n",
 937 |        "      <td>-0.050912</td>\n",
 938 |        "      <td>-0.989347</td>\n",
 939 |        "      <td>0.825994</td>\n",
 940 |        "      <td>0.949567</td>\n",
 941 |        "      <td>0.105761</td>\n",
 942 |        "      <td>-0.235078</td>\n",
 943 |        "      <td>0.986250</td>\n",
 944 |        "      <td>0.526325</td>\n",
 945 |        "      <td>-0.741661</td>\n",
 946 |        "      <td>0.819501</td>\n",
 947 |        "      <td>0.090216</td>\n",
 948 |        "      <td>-0.537516</td>\n",
 949 |        "      <td>0.846942</td>\n",
 950 |        "      <td>-0.385990</td>\n",
 951 |        "      <td>-0.578650</td>\n",
 952 |        "      <td>0.616461</td>\n",
 953 |        "      <td>0.490440</td>\n",
 954 |        "      <td>-0.002268</td>\n",
 955 |        "      <td>-0.768049</td>\n",
 956 |        "      <td>-0.745449</td>\n",
 957 |        "      <td>0.941508</td>\n",
 958 |        "      <td>0.847960</td>\n",
 959 |        "      <td>-0.947888</td>\n",
 960 |        "      <td>-0.421278</td>\n",
 961 |        "      <td>0.855621</td>\n",
 962 |        "      <td>-0.841248</td>\n",
 963 |        "      <td>-0.738750</td>\n",
 964 |        "      <td>-0.616005</td>\n",
 965 |        "      <td>0.000840</td>\n",
 966 |        "      <td>-0.962296</td>\n",
 967 |        "      <td>-0.913448</td>\n",
 968 |        "      <td>0.914193</td>\n",
 969 |        "      <td>-0.680750</td>\n",
 970 |        "      <td>-0.764314</td>\n",
 971 |        "      <td>-0.668780</td>\n",
 972 |        "      <td>0.553494</td>\n",
 973 |        "      <td>-0.927593</td>\n",
 974 |        "      <td>-0.647122</td>\n",
 975 |        "      <td>0.326478</td>\n",
 976 |        "      <td>0.968075</td>\n",
 977 |        "      <td>0.946287</td>\n",
 978 |        "      <td>-0.426579</td>\n",
 979 |        "      <td>0.674657</td>\n",
 980 |        "      <td>0.955783</td>\n",
 981 |        "      <td>-0.978836</td>\n",
 982 |        "      <td>0.512857</td>\n",
 983 |        "      <td>0.356463</td>\n",
 984 |        "      <td>-0.867495</td>\n",
 985 |        "      <td>0.154614</td>\n",
 986 |        "      <td>-0.958078</td>\n",
 987 |        "      <td>0.485347</td>\n",
 988 |        "      <td>0.984303</td>\n",
 989 |        "      <td>-0.659389</td>\n",
 990 |        "      <td>-0.967570</td>\n",
 991 |        "      <td>0.280092</td>\n",
 992 |        "      <td>-0.011816</td>\n",
 993 |        "      <td>-0.862490</td>\n",
 994 |        "      <td>-0.785080</td>\n",
 995 |        "      <td>0.212533</td>\n",
 996 |        "      <td>0.099250</td>\n",
 997 |        "      <td>0.970914</td>\n",
 998 |        "      <td>0.841250</td>\n",
 999 |        "      <td>0.772792</td>\n",
1000 |        "      <td>0.999893</td>\n",
1001 |        "      <td>-0.620373</td>\n",
1002 |        "      <td>-0.999964</td>\n",
1003 |        "      <td>-0.328710</td>\n",
1004 |        "      <td>0.549286</td>\n",
1005 |        "      <td>-0.995617</td>\n",
1006 |        "      <td>-0.654751</td>\n",
1007 |        "      <td>0.089768</td>\n",
1008 |        "      <td>0.955292</td>\n",
1009 |        "      <td>0.772715</td>\n",
1010 |        "      <td>-0.939837</td>\n",
1011 |        "    </tr>\n",
1012 |        "    <tr>\n",
1013 |        "      <th>3</th>\n",
1014 |        "      <td>ID_train_100</td>\n",
1015 |        "      <td>A</td>\n",
1016 |        "      <td>53.850238</td>\n",
1017 |        "      <td>NaN</td>\n",
1018 |        "      <td>NaN</td>\n",
1019 |        "      <td>NaN</td>\n",
1020 |        "      <td>NaN</td>\n",
1021 |        "      <td>NaN</td>\n",
1022 |        "      <td>22.966667</td>\n",
1023 |        "      <td>24.266667</td>\n",
1024 |        "      <td>25.275000</td>\n",
1025 |        "      <td>25.625000</td>\n",
1026 |        "      <td>25.866667</td>\n",
1027 |        "      <td>25.091667</td>\n",
1028 |        "      <td>24.025000</td>\n",
1029 |        "      <td>22.425000</td>\n",
1030 |        "      <td>20.233333</td>\n",
1031 |        "      <td>17.416667</td>\n",
1032 |        "      <td>17.391667</td>\n",
1033 |        "      <td>17.083333</td>\n",
1034 |        "      <td>17.516667</td>\n",
1035 |        "      <td>17.825000</td>\n",
1036 |        "      <td>18.258333</td>\n",
1037 |        "      <td>17.950000</td>\n",
1038 |        "      <td>17.575000</td>\n",
1039 |        "      <td>17.425000</td>\n",
1040 |        "      <td>17.475000</td>\n",
1041 |        "      <td>16.941667</td>\n",
1042 |        "      <td>18.233333</td>\n",
1043 |        "      <td>21.166667</td>\n",
1044 |        "      <td>23.708333</td>\n",
1045 |        "      <td>24.941667</td>\n",
1046 |        "      <td>25.825000</td>\n",
1047 |        "      <td>26.691667</td>\n",
1048 |        "      <td>27.275000</td>\n",
1049 |        "      <td>27.491667</td>\n",
1050 |        "      <td>21.136364</td>\n",
1051 |        "      <td>19.650000</td>\n",
1052 |        "      <td>19.391667</td>\n",
1053 |        "      <td>19.116667</td>\n",
1054 |        "      <td>19.075000</td>\n",
1055 |        "      <td>19.541667</td>\n",
1056 |        "      <td>19.450000</td>\n",
1057 |        "      <td>19.175000</td>\n",
1058 |        "      <td>18.891667</td>\n",
1059 |        "      <td>19.058333</td>\n",
1060 |        "      <td>19.325000</td>\n",
1061 |        "      <td>19.400000</td>\n",
1062 |        "      <td>19.458333</td>\n",
1063 |        "      <td>19.466667</td>\n",
1064 |        "      <td>19.441667</td>\n",
1065 |        "      <td>19.408333</td>\n",
1066 |        "      <td>19.850000</td>\n",
1067 |        "      <td>22.716667</td>\n",
1068 |        "      <td>23.091667</td>\n",
1069 |        "      <td>23.841667</td>\n",
1070 |        "      <td>25.166667</td>\n",
1071 |        "      <td>26.875000</td>\n",
1072 |        "      <td>27.125000</td>\n",
1073 |        "      <td>27.183333</td>\n",
1074 |        "      <td>26.500000</td>\n",
1075 |        "      <td>26.075000</td>\n",
1076 |        "      <td>23.391667</td>\n",
1077 |        "      <td>21.558333</td>\n",
1078 |        "      <td>20.525000</td>\n",
1079 |        "      <td>18.016667</td>\n",
1080 |        "      <td>18.333333</td>\n",
1081 |        "      <td>18.533333</td>\n",
1082 |        "      <td>18.641667</td>\n",
1083 |        "      <td>18.808333</td>\n",
1084 |        "      <td>18.741667</td>\n",
1085 |        "      <td>18.733333</td>\n",
1086 |        "      <td>18.575000</td>\n",
1087 |        "      <td>18.491667</td>\n",
1088 |        "      <td>18.225000</td>\n",
1089 |        "      <td>18.250000</td>\n",
1090 |        "      <td>18.458333</td>\n",
1091 |        "      <td>19.266667</td>\n",
1092 |        "      <td>21.025000</td>\n",
1093 |        "      <td>23.166667</td>\n",
1094 |        "      <td>24.000</td>\n",
1095 |        "      <td>24.708333</td>\n",
1096 |        "      <td>25.558333</td>\n",
1097 |        "      <td>26.500000</td>\n",
1098 |        "      <td>25.716667</td>\n",
1099 |        "      <td>20.758333</td>\n",
1100 |        "      <td>17.358333</td>\n",
1101 |        "      <td>18.116667</td>\n",
1102 |        "      <td>18.133333</td>\n",
1103 |        "      <td>18.058333</td>\n",
1104 |        "      <td>18.350000</td>\n",
1105 |        "      <td>18.266667</td>\n",
1106 |        "      <td>18.066667</td>\n",
1107 |        "      <td>18.133333</td>\n",
1108 |        "      <td>18.016667</td>\n",
1109 |        "      <td>17.450000</td>\n",
1110 |        "      <td>17.066667</td>\n",
1111 |        "      <td>17.325000</td>\n",
1112 |        "      <td>17.341667</td>\n",
1113 |        "      <td>17.375000</td>\n",
1114 |        "      <td>...</td>\n",
1115 |        "      <td>0.106538</td>\n",
1116 |        "      <td>-0.656670</td>\n",
1117 |        "      <td>0.127297</td>\n",
1118 |        "      <td>0.936974</td>\n",
1119 |        "      <td>-0.755199</td>\n",
1120 |        "      <td>0.225715</td>\n",
1121 |        "      <td>-0.876023</td>\n",
1122 |        "      <td>-0.780112</td>\n",
1123 |        "      <td>-0.930897</td>\n",
1124 |        "      <td>0.999544</td>\n",
1125 |        "      <td>-0.888784</td>\n",
1126 |        "      <td>-0.874682</td>\n",
1127 |        "      <td>-0.153198</td>\n",
1128 |        "      <td>0.534190</td>\n",
1129 |        "      <td>-0.293604</td>\n",
1130 |        "      <td>0.923470</td>\n",
1131 |        "      <td>-0.319239</td>\n",
1132 |        "      <td>-0.991811</td>\n",
1133 |        "      <td>-0.277354</td>\n",
1134 |        "      <td>-0.121781</td>\n",
1135 |        "      <td>-0.040652</td>\n",
1136 |        "      <td>-0.881798</td>\n",
1137 |        "      <td>-0.630439</td>\n",
1138 |        "      <td>0.606618</td>\n",
1139 |        "      <td>-0.811544</td>\n",
1140 |        "      <td>-0.991091</td>\n",
1141 |        "      <td>-0.999107</td>\n",
1142 |        "      <td>0.925944</td>\n",
1143 |        "      <td>0.997952</td>\n",
1144 |        "      <td>0.885999</td>\n",
1145 |        "      <td>0.684460</td>\n",
1146 |        "      <td>-0.414167</td>\n",
1147 |        "      <td>0.989378</td>\n",
1148 |        "      <td>-0.226863</td>\n",
1149 |        "      <td>0.045114</td>\n",
1150 |        "      <td>-0.104814</td>\n",
1151 |        "      <td>0.307146</td>\n",
1152 |        "      <td>-0.383696</td>\n",
1153 |        "      <td>-0.425937</td>\n",
1154 |        "      <td>0.429735</td>\n",
1155 |        "      <td>0.996446</td>\n",
1156 |        "      <td>-0.875655</td>\n",
1157 |        "      <td>0.925057</td>\n",
1158 |        "      <td>-0.823499</td>\n",
1159 |        "      <td>-0.208259</td>\n",
1160 |        "      <td>0.860431</td>\n",
1161 |        "      <td>0.047207</td>\n",
1162 |        "      <td>-0.357716</td>\n",
1163 |        "      <td>0.947255</td>\n",
1164 |        "      <td>-0.989517</td>\n",
1165 |        "      <td>0.777529</td>\n",
1166 |        "      <td>-0.200849</td>\n",
1167 |        "      <td>0.847853</td>\n",
1168 |        "      <td>0.269371</td>\n",
1169 |        "      <td>-0.584174</td>\n",
1170 |        "      <td>-0.657106</td>\n",
1171 |        "      <td>0.458151</td>\n",
1172 |        "      <td>-0.933249</td>\n",
1173 |        "      <td>0.824032</td>\n",
1174 |        "      <td>0.722134</td>\n",
1175 |        "      <td>0.991245</td>\n",
1176 |        "      <td>-0.613111</td>\n",
1177 |        "      <td>-0.168375</td>\n",
1178 |        "      <td>0.209887</td>\n",
1179 |        "      <td>0.349680</td>\n",
1180 |        "      <td>0.415632</td>\n",
1181 |        "      <td>0.890022</td>\n",
1182 |        "      <td>0.992949</td>\n",
1183 |        "      <td>0.875658</td>\n",
1184 |        "      <td>0.960588</td>\n",
1185 |        "      <td>0.202772</td>\n",
1186 |        "      <td>-0.415800</td>\n",
1187 |        "      <td>0.881197</td>\n",
1188 |        "      <td>0.178436</td>\n",
1189 |        "      <td>-0.713055</td>\n",
1190 |        "      <td>-0.977995</td>\n",
1191 |        "      <td>-0.868449</td>\n",
1192 |        "      <td>0.082654</td>\n",
1193 |        "      <td>-0.936302</td>\n",
1194 |        "      <td>0.559380</td>\n",
1195 |        "      <td>-0.926564</td>\n",
1196 |        "      <td>0.485755</td>\n",
1197 |        "      <td>1.000000</td>\n",
1198 |        "      <td>-0.202556</td>\n",
1199 |        "      <td>-0.498867</td>\n",
1200 |        "      <td>-0.267156</td>\n",
1201 |        "      <td>-0.901780</td>\n",
1202 |        "      <td>0.605642</td>\n",
1203 |        "      <td>0.994990</td>\n",
1204 |        "      <td>-0.552490</td>\n",
1205 |        "      <td>0.580466</td>\n",
1206 |        "      <td>0.945695</td>\n",
1207 |        "      <td>0.669056</td>\n",
1208 |        "      <td>-0.074219</td>\n",
1209 |        "      <td>0.816325</td>\n",
1210 |        "      <td>0.049529</td>\n",
1211 |        "      <td>0.357827</td>\n",
1212 |        "      <td>0.236227</td>\n",
1213 |        "      <td>0.301397</td>\n",
1214 |        "      <td>-0.949435</td>\n",
1215 |        "    </tr>\n",
1216 |        "    <tr>\n",
1217 |        "      <th>4</th>\n",
1218 |        "      <td>ID_train_1000</td>\n",
1219 |        "      <td>A</td>\n",
1220 |        "      <td>177.418750</td>\n",
1221 |        "      <td>NaN</td>\n",
1222 |        "      <td>NaN</td>\n",
1223 |        "      <td>NaN</td>\n",
1224 |        "      <td>NaN</td>\n",
1225 |        "      <td>NaN</td>\n",
1226 |        "      <td>21.875000</td>\n",
1227 |        "      <td>21.575000</td>\n",
1228 |        "      <td>21.525000</td>\n",
1229 |        "      <td>21.433333</td>\n",
1230 |        "      <td>20.508333</td>\n",
1231 |        "      <td>19.916667</td>\n",
1232 |        "      <td>18.991667</td>\n",
1233 |        "      <td>18.591667</td>\n",
1234 |        "      <td>18.050000</td>\n",
1235 |        "      <td>17.916667</td>\n",
1236 |        "      <td>18.166667</td>\n",
1237 |        "      <td>17.525000</td>\n",
1238 |        "      <td>19.191667</td>\n",
1239 |        "      <td>22.458333</td>\n",
1240 |        "      <td>25.016667</td>\n",
1241 |        "      <td>25.858333</td>\n",
1242 |        "      <td>26.850000</td>\n",
1243 |        "      <td>27.625000</td>\n",
1244 |        "      <td>28.358333</td>\n",
1245 |        "      <td>28.750000</td>\n",
1246 |        "      <td>27.850000</td>\n",
1247 |        "      <td>23.766667</td>\n",
1248 |        "      <td>20.891667</td>\n",
1249 |        "      <td>21.008333</td>\n",
1250 |        "      <td>20.608333</td>\n",
1251 |        "      <td>20.191667</td>\n",
1252 |        "      <td>20.108333</td>\n",
1253 |        "      <td>19.633333</td>\n",
1254 |        "      <td>24.247348</td>\n",
1255 |        "      <td>23.537500</td>\n",
1256 |        "      <td>23.416667</td>\n",
1257 |        "      <td>22.900000</td>\n",
1258 |        "      <td>22.100000</td>\n",
1259 |        "      <td>21.554167</td>\n",
1260 |        "      <td>21.116667</td>\n",
1261 |        "      <td>20.970833</td>\n",
1262 |        "      <td>20.691667</td>\n",
1263 |        "      <td>20.587500</td>\n",
1264 |        "      <td>20.820833</td>\n",
1265 |        "      <td>21.108333</td>\n",
1266 |        "      <td>20.829167</td>\n",
1267 |        "      <td>20.879167</td>\n",
1268 |        "      <td>20.625000</td>\n",
1269 |        "      <td>18.608333</td>\n",
1270 |        "      <td>18.516667</td>\n",
1271 |        "      <td>20.020833</td>\n",
1272 |        "      <td>20.225000</td>\n",
1273 |        "      <td>20.612500</td>\n",
1274 |        "      <td>21.675000</td>\n",
1275 |        "      <td>22.925000</td>\n",
1276 |        "      <td>23.857955</td>\n",
1277 |        "      <td>25.779167</td>\n",
1278 |        "      <td>21.263636</td>\n",
1279 |        "      <td>20.616667</td>\n",
1280 |        "      <td>24.316667</td>\n",
1281 |        "      <td>23.008333</td>\n",
1282 |        "      <td>22.145833</td>\n",
1283 |        "      <td>20.095833</td>\n",
1284 |        "      <td>19.970833</td>\n",
1285 |        "      <td>20.083333</td>\n",
1286 |        "      <td>20.062500</td>\n",
1287 |        "      <td>19.933333</td>\n",
1288 |        "      <td>19.691667</td>\n",
1289 |        "      <td>19.816667</td>\n",
1290 |        "      <td>19.433333</td>\n",
1291 |        "      <td>19.441667</td>\n",
1292 |        "      <td>19.154167</td>\n",
1293 |        "      <td>19.016667</td>\n",
1294 |        "      <td>19.016667</td>\n",
1295 |        "      <td>19.525000</td>\n",
1296 |        "      <td>21.487500</td>\n",
1297 |        "      <td>22.937500</td>\n",
1298 |        "      <td>23.400</td>\n",
1299 |        "      <td>24.316667</td>\n",
1300 |        "      <td>25.304167</td>\n",
1301 |        "      <td>26.208333</td>\n",
1302 |        "      <td>26.145833</td>\n",
1303 |        "      <td>23.670833</td>\n",
1304 |        "      <td>21.725000</td>\n",
1305 |        "      <td>21.837500</td>\n",
1306 |        "      <td>21.279167</td>\n",
1307 |        "      <td>20.570833</td>\n",
1308 |        "      <td>20.387500</td>\n",
1309 |        "      <td>19.991667</td>\n",
1310 |        "      <td>19.833333</td>\n",
1311 |        "      <td>19.245833</td>\n",
1312 |        "      <td>18.691667</td>\n",
1313 |        "      <td>18.383333</td>\n",
1314 |        "      <td>18.325000</td>\n",
1315 |        "      <td>18.433333</td>\n",
1316 |        "      <td>18.437500</td>\n",
1317 |        "      <td>18.541667</td>\n",
1318 |        "      <td>...</td>\n",
1319 |        "      <td>-0.996037</td>\n",
1320 |        "      <td>0.763370</td>\n",
1321 |        "      <td>0.632792</td>\n",
1322 |        "      <td>-0.142443</td>\n",
1323 |        "      <td>0.883681</td>\n",
1324 |        "      <td>0.967957</td>\n",
1325 |        "      <td>0.978257</td>\n",
1326 |        "      <td>0.176707</td>\n",
1327 |        "      <td>0.989523</td>\n",
1328 |        "      <td>0.881470</td>\n",
1329 |        "      <td>-0.928733</td>\n",
1330 |        "      <td>0.166814</td>\n",
1331 |        "      <td>-0.959935</td>\n",
1332 |        "      <td>-0.936613</td>\n",
1333 |        "      <td>0.054542</td>\n",
1334 |        "      <td>-0.414697</td>\n",
1335 |        "      <td>-0.866427</td>\n",
1336 |        "      <td>0.173500</td>\n",
1337 |        "      <td>0.620190</td>\n",
1338 |        "      <td>-0.906068</td>\n",
1339 |        "      <td>-0.035412</td>\n",
1340 |        "      <td>0.999959</td>\n",
1341 |        "      <td>-0.830783</td>\n",
1342 |        "      <td>-0.269391</td>\n",
1343 |        "      <td>0.114415</td>\n",
1344 |        "      <td>0.054568</td>\n",
1345 |        "      <td>-0.071428</td>\n",
1346 |        "      <td>-0.999393</td>\n",
1347 |        "      <td>-0.916862</td>\n",
1348 |        "      <td>0.966995</td>\n",
1349 |        "      <td>0.827238</td>\n",
1350 |        "      <td>-0.550644</td>\n",
1351 |        "      <td>-0.426123</td>\n",
1352 |        "      <td>-0.532556</td>\n",
1353 |        "      <td>-0.587063</td>\n",
1354 |        "      <td>-0.920966</td>\n",
1355 |        "      <td>-0.655409</td>\n",
1356 |        "      <td>-0.218329</td>\n",
1357 |        "      <td>-0.129844</td>\n",
1358 |        "      <td>0.959347</td>\n",
1359 |        "      <td>-0.707344</td>\n",
1360 |        "      <td>-0.337041</td>\n",
1361 |        "      <td>0.961337</td>\n",
1362 |        "      <td>-0.935177</td>\n",
1363 |        "      <td>0.440540</td>\n",
1364 |        "      <td>-0.942829</td>\n",
1365 |        "      <td>-0.969093</td>\n",
1366 |        "      <td>-0.985078</td>\n",
1367 |        "      <td>-0.946619</td>\n",
1368 |        "      <td>-0.662913</td>\n",
1369 |        "      <td>-0.327247</td>\n",
1370 |        "      <td>-0.979234</td>\n",
1371 |        "      <td>0.738470</td>\n",
1372 |        "      <td>0.558170</td>\n",
1373 |        "      <td>0.496131</td>\n",
1374 |        "      <td>-0.045062</td>\n",
1375 |        "      <td>0.861434</td>\n",
1376 |        "      <td>0.406506</td>\n",
1377 |        "      <td>0.440201</td>\n",
1378 |        "      <td>-0.677080</td>\n",
1379 |        "      <td>-0.840071</td>\n",
1380 |        "      <td>0.669422</td>\n",
1381 |        "      <td>0.750988</td>\n",
1382 |        "      <td>0.277206</td>\n",
1383 |        "      <td>0.372669</td>\n",
1384 |        "      <td>-0.823120</td>\n",
1385 |        "      <td>0.290495</td>\n",
1386 |        "      <td>-0.201678</td>\n",
1387 |        "      <td>-0.264323</td>\n",
1388 |        "      <td>0.965121</td>\n",
1389 |        "      <td>0.817075</td>\n",
1390 |        "      <td>-0.984789</td>\n",
1391 |        "      <td>0.975569</td>\n",
1392 |        "      <td>0.637579</td>\n",
1393 |        "      <td>-0.985222</td>\n",
1394 |        "      <td>-0.991592</td>\n",
1395 |        "      <td>0.728332</td>\n",
1396 |        "      <td>-0.008841</td>\n",
1397 |        "      <td>-0.205911</td>\n",
1398 |        "      <td>0.967486</td>\n",
1399 |        "      <td>0.379974</td>\n",
1400 |        "      <td>-0.949185</td>\n",
1401 |        "      <td>0.983956</td>\n",
1402 |        "      <td>-0.404907</td>\n",
1403 |        "      <td>-0.988736</td>\n",
1404 |        "      <td>-0.293643</td>\n",
1405 |        "      <td>0.981650</td>\n",
1406 |        "      <td>-0.891477</td>\n",
1407 |        "      <td>0.962301</td>\n",
1408 |        "      <td>-0.941072</td>\n",
1409 |        "      <td>0.461160</td>\n",
1410 |        "      <td>0.970801</td>\n",
1411 |        "      <td>-0.165447</td>\n",
1412 |        "      <td>-0.791261</td>\n",
1413 |        "      <td>-0.901593</td>\n",
1414 |        "      <td>-0.491141</td>\n",
1415 |        "      <td>0.996079</td>\n",
1416 |        "      <td>-0.391008</td>\n",
1417 |        "      <td>-0.302821</td>\n",
1418 |        "      <td>-0.618876</td>\n",
1419 |        "    </tr>\n",
1420 |        "  </tbody>\n",
1421 |        "</table>\n",
1422 |        "<p>5 rows × 857 columns</p>\n",
1423 |        "</div>"
1424 |       ],
1425 |       "text/plain": [
1426 |        "              ID location      target  temp  precip  rel_humidity  wind_spd  \\\n",
1427 |        "0     ID_train_0        C   45.126304   NaN     NaN           NaN       NaN   \n",
1428 |        "1     ID_train_1        D   79.131702   NaN     NaN           NaN       NaN   \n",
1429 |        "2    ID_train_10        A   32.661304   NaN     NaN           NaN       NaN   \n",
1430 |        "3   ID_train_100        A   53.850238   NaN     NaN           NaN       NaN   \n",
1431 |        "4  ID_train_1000        A  177.418750   NaN     NaN           NaN       NaN   \n",
1432 |        "\n",
1433 |        "   atmos_press     temp_0     temp_1     temp_2     temp_3     temp_4  \\\n",
1434 |        "0          NaN  22.533333  21.716667  20.833333  20.983333  20.875000   \n",
1435 |        "1          NaN  22.533333  21.716667  20.833333  20.983333  20.875000   \n",
1436 |        "2          NaN  28.975000  27.950000  29.600000  26.425000  22.091667   \n",
1437 |        "3          NaN  22.966667  24.266667  25.275000  25.625000  25.866667   \n",
1438 |        "4          NaN  21.875000  21.575000  21.525000  21.433333  20.508333   \n",
1439 |        "\n",
1440 |        "      temp_5     temp_6     temp_7     temp_8     temp_9    temp_10  \\\n",
1441 |        "0  20.141667  19.375000  18.791667  18.775000  18.641667  18.558333   \n",
1442 |        "1  20.141667  19.375000  18.791667  18.775000  18.641667  18.558333   \n",
1443 |        "2  21.775000  22.333333  21.950000  20.441667  20.441667  20.950000   \n",
1444 |        "3  25.091667  24.025000  22.425000  20.233333  17.416667  17.391667   \n",
1445 |        "4  19.916667  18.991667  18.591667  18.050000  17.916667  18.166667   \n",
1446 |        "\n",
1447 |        "     temp_11    temp_12    temp_13    temp_14    temp_15    temp_16  \\\n",
1448 |        "0  18.533333  19.608333  21.916667  24.716667  26.658333  29.175000   \n",
1449 |        "1  18.533333  19.608333  21.916667  24.716667  26.658333  29.175000   \n",
1450 |        "2  19.800000  19.591667  19.575000  19.516667  19.550000  19.783333   \n",
1451 |        "3  17.083333  17.516667  17.825000  18.258333  17.950000  17.575000   \n",
1452 |        "4  17.525000  19.191667  22.458333  25.016667  25.858333  26.850000   \n",
1453 |        "\n",
1454 |        "     temp_17    temp_18    temp_19    temp_20    temp_21    temp_22  \\\n",
1455 |        "0  30.700000  31.433333  32.333333  33.008333  33.391667  33.616667   \n",
1456 |        "1  30.700000  31.433333  32.333333  33.008333  33.391667  33.616667   \n",
1457 |        "2  19.908333  19.516667  19.666667  20.683333  22.491667  23.708333   \n",
1458 |        "3  17.425000  17.475000  16.941667  18.233333  21.166667  23.708333   \n",
1459 |        "4  27.625000  28.358333  28.750000  27.850000  23.766667  20.891667   \n",
1460 |        "\n",
1461 |        "     temp_23    temp_24    temp_25    temp_26    temp_27    temp_28  \\\n",
1462 |        "0  31.091667  27.550000  26.658333  25.675000  26.425000  23.783333   \n",
1463 |        "1  31.091667  27.550000  26.658333  25.675000  26.425000  23.783333   \n",
1464 |        "2  24.900000  26.058333  27.325000  27.866667  28.291667  22.136364   \n",
1465 |        "3  24.941667  25.825000  26.691667  27.275000  27.491667  21.136364   \n",
1466 |        "4  21.008333  20.608333  20.191667  20.108333  19.633333  24.247348   \n",
1467 |        "\n",
1468 |        "     temp_29    temp_30    temp_31    temp_32    temp_33    temp_34  \\\n",
1469 |        "0  22.416667  22.041667  21.008333  21.475000  21.825000  21.158333   \n",
1470 |        "1  22.416667  22.041667  21.008333  21.475000  21.825000  21.158333   \n",
1471 |        "2  19.141667  19.991667  20.216667  20.433333  20.466667  20.800000   \n",
1472 |        "3  19.650000  19.391667  19.116667  19.075000  19.541667  19.450000   \n",
1473 |        "4  23.537500  23.416667  22.900000  22.100000  21.554167  21.116667   \n",
1474 |        "\n",
1475 |        "     temp_35    temp_36    temp_37    temp_38    temp_39    temp_40  \\\n",
1476 |        "0  22.308333  23.116667  22.333333  23.850000  23.825000  24.983333   \n",
1477 |        "1  22.308333  23.116667  22.333333  23.850000  23.825000  24.983333   \n",
1478 |        "2  20.958333  20.641667  20.033333  19.825000  19.616667  19.225000   \n",
1479 |        "3  19.175000  18.891667  19.058333  19.325000  19.400000  19.458333   \n",
1480 |        "4  20.970833  20.691667  20.587500  20.820833  21.108333  20.829167   \n",
1481 |        "\n",
1482 |        "     temp_41    temp_42    temp_43    temp_44    temp_45    temp_46  \\\n",
1483 |        "0  28.050000  30.841667  31.991667  32.041667  32.000000  31.666667   \n",
1484 |        "1  28.050000  30.841667  31.991667  32.041667  32.000000  31.666667   \n",
1485 |        "2  18.716667  18.458333  19.000000  20.758333  27.358333  24.491667   \n",
1486 |        "3  19.466667  19.441667  19.408333  19.850000  22.716667  23.091667   \n",
1487 |        "4  20.879167  20.625000  18.608333  18.516667  20.020833  20.225000   \n",
1488 |        "\n",
1489 |        "     temp_47    temp_48    temp_49    temp_50    temp_51    temp_52  \\\n",
1490 |        "0  29.158333  25.866667  24.300000  23.683333  23.758333  22.991667   \n",
1491 |        "1  29.158333  25.866667  24.300000  23.683333  23.758333  22.991667   \n",
1492 |        "2  26.350000  27.950000  29.166667  30.066667  30.550000  30.400000   \n",
1493 |        "3  23.841667  25.166667  26.875000  27.125000  27.183333  26.500000   \n",
1494 |        "4  20.612500  21.675000  22.925000  23.857955  25.779167  21.263636   \n",
1495 |        "\n",
1496 |        "     temp_53    temp_54    temp_55    temp_56    temp_57    temp_58  \\\n",
1497 |        "0  21.891667  21.158333  20.416667  19.533333  19.125000  18.766667   \n",
1498 |        "1  21.891667  21.158333  20.416667  19.533333  19.125000  18.766667   \n",
1499 |        "2  30.008333  28.908333  26.700000  25.716667  24.858333  24.291667   \n",
1500 |        "3  26.075000  23.391667  21.558333  20.525000  18.016667  18.333333   \n",
1501 |        "4  20.616667  24.316667  23.008333  22.145833  20.095833  19.970833   \n",
1502 |        "\n",
1503 |        "     temp_59    temp_60    temp_61    temp_62    temp_63    temp_64  \\\n",
1504 |        "0  17.983333  18.458333  21.933333  23.791667  25.558333  27.758333   \n",
1505 |        "1  17.983333  18.458333  21.933333  23.791667  25.558333  27.758333   \n",
1506 |        "2  23.308333  23.008333  22.041667  21.375000  21.133333  21.158333   \n",
1507 |        "3  18.533333  18.641667  18.808333  18.741667  18.733333  18.575000   \n",
1508 |        "4  20.083333  20.062500  19.933333  19.691667  19.816667  19.433333   \n",
1509 |        "\n",
1510 |        "     temp_65    temp_66    temp_67    temp_68    temp_69    temp_70  \\\n",
1511 |        "0  29.658333  31.391667  31.883333  32.358333  32.708333  31.983333   \n",
1512 |        "1  29.658333  31.391667  31.883333  32.358333  32.708333  31.983333   \n",
1513 |        "2  20.658333  20.766667  21.633333  22.708333  24.808333  27.108333   \n",
1514 |        "3  18.491667  18.225000  18.250000  18.458333  19.266667  21.025000   \n",
1515 |        "4  19.441667  19.154167  19.016667  19.016667  19.525000  21.487500   \n",
1516 |        "\n",
1517 |        "     temp_71  temp_72    temp_73    temp_74    temp_75    temp_76    temp_77  \\\n",
1518 |        "0  30.850000   28.800  26.491667  24.866667  24.366667  23.025000  22.325000   \n",
1519 |        "1  30.850000   28.800  26.491667  24.866667  24.366667  23.025000  22.325000   \n",
1520 |        "2  28.775000   29.475  29.766667  29.875000  29.150000  27.716667  27.491667   \n",
1521 |        "3  23.166667   24.000  24.708333  25.558333  26.500000  25.716667  20.758333   \n",
1522 |        "4  22.937500   23.400  24.316667  25.304167  26.208333  26.145833  23.670833   \n",
1523 |        "\n",
1524 |        "     temp_78    temp_79    temp_80    temp_81    temp_82    temp_83  \\\n",
1525 |        "0  21.650000  20.750000  20.475000  19.641667  19.516667  19.575000   \n",
1526 |        "1  21.650000  20.750000  20.475000  19.641667  19.516667  19.575000   \n",
1527 |        "2  26.483333  25.475000  25.008333  24.600000  24.033333  23.358333   \n",
1528 |        "3  17.358333  18.116667  18.133333  18.058333  18.350000  18.266667   \n",
1529 |        "4  21.725000  21.837500  21.279167  20.570833  20.387500  19.991667   \n",
1530 |        "\n",
1531 |        "     temp_84    temp_85    temp_86    temp_87    temp_88    temp_89  \\\n",
1532 |        "0  20.000000  23.358333  25.608333  26.883333  26.358333  27.325000   \n",
1533 |        "1  20.000000  23.358333  25.608333  26.883333  26.358333  27.325000   \n",
1534 |        "2  22.366667  22.608333  22.741667  21.908333  21.550000  21.758333   \n",
1535 |        "3  18.066667  18.133333  18.016667  17.450000  17.066667  17.325000   \n",
1536 |        "4  19.833333  19.245833  18.691667  18.383333  18.325000  18.433333   \n",
1537 |        "\n",
1538 |        "     temp_90    temp_91  ...  wind_dir_sin_21  wind_dir_sin_22  \\\n",
1539 |        "0  29.008333  28.433333  ...        -0.141620         0.677692   \n",
1540 |        "1  29.008333  28.433333  ...        -0.141620         0.677692   \n",
1541 |        "2  20.316667  20.650000  ...         0.992270        -0.744836   \n",
1542 |        "3  17.341667  17.375000  ...         0.106538        -0.656670   \n",
1543 |        "4  18.437500  18.541667  ...        -0.996037         0.763370   \n",
1544 |        "\n",
1545 |        "   wind_dir_sin_23  wind_dir_sin_24  wind_dir_sin_25  wind_dir_sin_26  \\\n",
1546 |        "0        -0.691042        -0.570434        -0.621553         0.054916   \n",
1547 |        "1        -0.691042        -0.570434        -0.621553         0.054916   \n",
1548 |        "2         0.312328        -0.613390        -0.691339        -0.996273   \n",
1549 |        "3         0.127297         0.936974        -0.755199         0.225715   \n",
1550 |        "4         0.632792        -0.142443         0.883681         0.967957   \n",
1551 |        "\n",
1552 |        "   wind_dir_sin_27  wind_dir_sin_28  wind_dir_sin_29  wind_dir_sin_30  \\\n",
1553 |        "0        -0.068411        -0.427471        -0.748119         0.795807   \n",
1554 |        "1        -0.068411        -0.427471        -0.748119         0.795807   \n",
1555 |        "2         0.877801        -0.619629        -0.852653         0.952294   \n",
1556 |        "3        -0.876023        -0.780112        -0.930897         0.999544   \n",
1557 |        "4         0.978257         0.176707         0.989523         0.881470   \n",
1558 |        "\n",
1559 |        "   wind_dir_sin_31  wind_dir_sin_32  wind_dir_sin_33  wind_dir_sin_34  \\\n",
1560 |        "0         0.997073        -0.105681        -0.170296         0.850561   \n",
1561 |        "1         0.997073        -0.105681        -0.170296         0.850561   \n",
1562 |        "2         0.272820         0.887248         0.161753         0.710564   \n",
1563 |        "3        -0.888784        -0.874682        -0.153198         0.534190   \n",
1564 |        "4        -0.928733         0.166814        -0.959935        -0.936613   \n",
1565 |        "\n",
1566 |        "   wind_dir_sin_35  wind_dir_sin_36  wind_dir_sin_37  wind_dir_sin_38  \\\n",
1567 |        "0         0.971100        -0.002620        -0.897881         0.855572   \n",
1568 |        "1         0.971100        -0.002620        -0.897881         0.855572   \n",
1569 |        "2         0.493320         0.831038         0.922995        -0.206542   \n",
1570 |        "3        -0.293604         0.923470        -0.319239        -0.991811   \n",
1571 |        "4         0.054542        -0.414697        -0.866427         0.173500   \n",
1572 |        "\n",
1573 |        "   wind_dir_sin_39  wind_dir_sin_40  wind_dir_sin_41  wind_dir_sin_42  \\\n",
1574 |        "0         0.562276         0.606049         0.902810        -0.853162   \n",
1575 |        "1         0.562276         0.606049         0.902810        -0.853162   \n",
1576 |        "2        -0.905140         0.261427        -0.834561         0.999585   \n",
1577 |        "3        -0.277354        -0.121781        -0.040652        -0.881798   \n",
1578 |        "4         0.620190        -0.906068        -0.035412         0.999959   \n",
1579 |        "\n",
1580 |        "   wind_dir_sin_43  wind_dir_sin_44  wind_dir_sin_45  wind_dir_sin_46  \\\n",
1581 |        "0        -0.186615        -0.032798         0.937824        -0.317277   \n",
1582 |        "1        -0.186615        -0.032798         0.937824        -0.317277   \n",
1583 |        "2         0.969601        -0.304580        -0.134351         0.926241   \n",
1584 |        "3        -0.630439         0.606618        -0.811544        -0.991091   \n",
1585 |        "4        -0.830783        -0.269391         0.114415         0.054568   \n",
1586 |        "\n",
1587 |        "   wind_dir_sin_47  wind_dir_sin_48  wind_dir_sin_49  wind_dir_sin_50  \\\n",
1588 |        "0        -0.451505        -0.797416        -0.159037        -0.842704   \n",
1589 |        "1        -0.451505        -0.797416        -0.159037        -0.842704   \n",
1590 |        "2        -0.050912        -0.989347         0.825994         0.949567   \n",
1591 |        "3        -0.999107         0.925944         0.997952         0.885999   \n",
1592 |        "4        -0.071428        -0.999393        -0.916862         0.966995   \n",
1593 |        "\n",
1594 |        "   wind_dir_sin_51  wind_dir_sin_52  wind_dir_sin_53  wind_dir_sin_54  \\\n",
1595 |        "0         0.590000         0.249485         0.643785         0.369674   \n",
1596 |        "1         0.590000         0.249485         0.643785         0.369674   \n",
1597 |        "2         0.105761        -0.235078         0.986250         0.526325   \n",
1598 |        "3         0.684460        -0.414167         0.989378        -0.226863   \n",
1599 |        "4         0.827238        -0.550644        -0.426123        -0.532556   \n",
1600 |        "\n",
1601 |        "   wind_dir_sin_55  wind_dir_sin_56  wind_dir_sin_57  wind_dir_sin_58  \\\n",
1602 |        "0        -0.910274        -0.390897         0.531411        -0.117671   \n",
1603 |        "1        -0.910274        -0.390897         0.531411        -0.117671   \n",
1604 |        "2        -0.741661         0.819501         0.090216        -0.537516   \n",
1605 |        "3         0.045114        -0.104814         0.307146        -0.383696   \n",
1606 |        "4        -0.587063        -0.920966        -0.655409        -0.218329   \n",
1607 |        "\n",
1608 |        "   wind_dir_sin_59  wind_dir_sin_60  wind_dir_sin_61  wind_dir_sin_62  \\\n",
1609 |        "0         0.682072         0.997411        -0.445040         0.641554   \n",
1610 |        "1         0.682072         0.997411        -0.445040         0.641554   \n",
1611 |        "2         0.846942        -0.385990        -0.578650         0.616461   \n",
1612 |        "3        -0.425937         0.429735         0.996446        -0.875655   \n",
1613 |        "4        -0.129844         0.959347        -0.707344        -0.337041   \n",
1614 |        "\n",
1615 |        "   wind_dir_sin_63  wind_dir_sin_64  wind_dir_sin_65  wind_dir_sin_66  \\\n",
1616 |        "0         0.917998         0.647900         0.843400         0.831011   \n",
1617 |        "1         0.917998         0.647900         0.843400         0.831011   \n",
1618 |        "2         0.490440        -0.002268        -0.768049        -0.745449   \n",
1619 |        "3         0.925057        -0.823499        -0.208259         0.860431   \n",
1620 |        "4         0.961337        -0.935177         0.440540        -0.942829   \n",
1621 |        "\n",
1622 |        "   wind_dir_sin_67  wind_dir_sin_68  wind_dir_sin_69  wind_dir_sin_70  \\\n",
1623 |        "0        -0.780758        -0.402907        -0.970055        -0.775132   \n",
1624 |        "1        -0.780758        -0.402907        -0.970055        -0.775132   \n",
1625 |        "2         0.941508         0.847960        -0.947888        -0.421278   \n",
1626 |        "3         0.047207        -0.357716         0.947255        -0.989517   \n",
1627 |        "4        -0.969093        -0.985078        -0.946619        -0.662913   \n",
1628 |        "\n",
1629 |        "   wind_dir_sin_71  wind_dir_sin_72  wind_dir_sin_73  wind_dir_sin_74  \\\n",
1630 |        "0        -0.737531        -0.877086         0.706791        -0.991253   \n",
1631 |        "1        -0.737531        -0.877086         0.706791        -0.991253   \n",
1632 |        "2         0.855621        -0.841248        -0.738750        -0.616005   \n",
1633 |        "3         0.777529        -0.200849         0.847853         0.269371   \n",
1634 |        "4        -0.327247        -0.979234         0.738470         0.558170   \n",
1635 |        "\n",
1636 |        "   wind_dir_sin_75  wind_dir_sin_76  wind_dir_sin_77  wind_dir_sin_78  \\\n",
1637 |        "0         0.870372         0.235040        -0.399780        -0.890181   \n",
1638 |        "1         0.870372         0.235040        -0.399780        -0.890181   \n",
1639 |        "2         0.000840        -0.962296        -0.913448         0.914193   \n",
1640 |        "3        -0.584174        -0.657106         0.458151        -0.933249   \n",
1641 |        "4         0.496131        -0.045062         0.861434         0.406506   \n",
1642 |        "\n",
1643 |        "   wind_dir_sin_79  wind_dir_sin_80  wind_dir_sin_81  wind_dir_sin_82  \\\n",
1644 |        "0        -0.429068         0.536505         0.849812        -0.921784   \n",
1645 |        "1        -0.429068         0.536505         0.849812        -0.921784   \n",
1646 |        "2        -0.680750        -0.764314        -0.668780         0.553494   \n",
1647 |        "3         0.824032         0.722134         0.991245        -0.613111   \n",
1648 |        "4         0.440201        -0.677080        -0.840071         0.669422   \n",
1649 |        "\n",
1650 |        "   wind_dir_sin_83  wind_dir_sin_84  wind_dir_sin_85  wind_dir_sin_86  \\\n",
1651 |        "0         0.890076         0.325722        -0.968946         0.422955   \n",
1652 |        "1         0.890076         0.325722        -0.968946         0.422955   \n",
1653 |        "2        -0.927593        -0.647122         0.326478         0.968075   \n",
1654 |        "3        -0.168375         0.209887         0.349680         0.415632   \n",
1655 |        "4         0.750988         0.277206         0.372669        -0.823120   \n",
1656 |        "\n",
1657 |        "   wind_dir_sin_87  wind_dir_sin_88  wind_dir_sin_89  wind_dir_sin_90  \\\n",
1658 |        "0        -0.908680        -0.238531         0.994556        -0.784712   \n",
1659 |        "1        -0.908680        -0.238531         0.994556        -0.784712   \n",
1660 |        "2         0.946287        -0.426579         0.674657         0.955783   \n",
1661 |        "3         0.890022         0.992949         0.875658         0.960588   \n",
1662 |        "4         0.290495        -0.201678        -0.264323         0.965121   \n",
1663 |        "\n",
1664 |        "   wind_dir_sin_91  wind_dir_sin_92  wind_dir_sin_93  wind_dir_sin_94  \\\n",
1665 |        "0        -0.327369         0.793215         0.921165         0.900036   \n",
1666 |        "1        -0.327369         0.793215         0.921165         0.900036   \n",
1667 |        "2        -0.978836         0.512857         0.356463        -0.867495   \n",
1668 |        "3         0.202772        -0.415800         0.881197         0.178436   \n",
1669 |        "4         0.817075        -0.984789         0.975569         0.637579   \n",
1670 |        "\n",
1671 |        "   wind_dir_sin_95  wind_dir_sin_96  wind_dir_sin_97  wind_dir_sin_98  \\\n",
1672 |        "0        -0.244630        -0.858170        -0.980876         0.890335   \n",
1673 |        "1        -0.244630        -0.858170        -0.980876         0.890335   \n",
1674 |        "2         0.154614        -0.958078         0.485347         0.984303   \n",
1675 |        "3        -0.713055        -0.977995        -0.868449         0.082654   \n",
1676 |        "4        -0.985222        -0.991592         0.728332        -0.008841   \n",
1677 |        "\n",
1678 |        "   wind_dir_sin_99  wind_dir_sin_100  wind_dir_sin_101  wind_dir_sin_102  \\\n",
1679 |        "0        -0.992381         -0.861434         -0.629305         -0.290513   \n",
1680 |        "1        -0.992381         -0.861434         -0.629305         -0.290513   \n",
1681 |        "2        -0.659389         -0.967570          0.280092         -0.011816   \n",
1682 |        "3        -0.936302          0.559380         -0.926564          0.485755   \n",
1683 |        "4        -0.205911          0.967486          0.379974         -0.949185   \n",
1684 |        "\n",
1685 |        "   wind_dir_sin_103  wind_dir_sin_104  wind_dir_sin_105  wind_dir_sin_106  \\\n",
1686 |        "0          0.999024         -0.985627          0.403930         -0.778758   \n",
1687 |        "1          0.999024         -0.985627          0.923642         -0.553773   \n",
1688 |        "2         -0.862490         -0.785080          0.212533          0.099250   \n",
1689 |        "3          1.000000         -0.202556         -0.498867         -0.267156   \n",
1690 |        "4          0.983956         -0.404907         -0.988736         -0.293643   \n",
1691 |        "\n",
1692 |        "   wind_dir_sin_107  wind_dir_sin_108  wind_dir_sin_109  wind_dir_sin_110  \\\n",
1693 |        "0         -0.623189          0.921356         -0.949976         -0.938622   \n",
1694 |        "1          0.455011         -0.989001          0.211568         -0.230857   \n",
1695 |        "2          0.970914          0.841250          0.772792          0.999893   \n",
1696 |        "3         -0.901780          0.605642          0.994990         -0.552490   \n",
1697 |        "4          0.981650         -0.891477          0.962301         -0.941072   \n",
1698 |        "\n",
1699 |        "   wind_dir_sin_111  wind_dir_sin_112  wind_dir_sin_113  wind_dir_sin_114  \\\n",
1700 |        "0          0.539989          0.833571         -0.566473          0.249451   \n",
1701 |        "1          0.971538          0.669779         -0.839579         -0.212372   \n",
1702 |        "2         -0.620373         -0.999964         -0.328710          0.549286   \n",
1703 |        "3          0.580466          0.945695          0.669056         -0.074219   \n",
1704 |        "4          0.461160          0.970801         -0.165447         -0.791261   \n",
1705 |        "\n",
1706 |        "   wind_dir_sin_115  wind_dir_sin_116  wind_dir_sin_117  wind_dir_sin_118  \\\n",
1707 |        "0          0.703657         -0.775581          0.150566          0.117369   \n",
1708 |        "1         -0.029492          0.903239          0.881671          0.304360   \n",
1709 |        "2         -0.995617         -0.654751          0.089768          0.955292   \n",
1710 |        "3          0.816325          0.049529          0.357827          0.236227   \n",
1711 |        "4         -0.901593         -0.491141          0.996079         -0.391008   \n",
1712 |        "\n",
1713 |        "   wind_dir_sin_119  wind_dir_sin_120  \n",
1714 |        "0          0.635717         -0.947955  \n",
1715 |        "1         -0.955722          0.996240  \n",
1716 |        "2          0.772715         -0.939837  \n",
1717 |        "3          0.301397         -0.949435  \n",
1718 |        "4         -0.302821         -0.618876  \n",
1719 |        "\n",
1720 |        "[5 rows x 857 columns]"
1721 |       ]
1722 |      },
1723 |      "execution_count": 8,
1724 |      "metadata": {},
1725 |      "output_type": "execute_result"
1726 |     }
1727 |    ],
1728 |    "source": [
1729 |     "%%time\n",
1730 |     "### Only takes 30s to preprocess the data\n",
1731 |     "data = pd.concat([train,test]).reset_index(drop=True)\n",
1732 |     "df = data[['ID', 'location', 'target']]\n",
1733 |     "df = pd.concat([train,test]).reset_index(drop=True)\n",
1734 |     "for c in tqdm_notebook(['temp', 'precip', 'rel_humidity', 'wind_dir','wind_spd', 'atmos_press']):\n",
1735 |     "    tmp = data[c].str.split(',', expand=True)\n",
1736 |     "    tmp.columns = [c + '_' + str(x) for x in tmp.columns]\n",
1737 |     "    df = pd.concat([df, tmp], axis=1)\n",
1738 |     "obj_cols = [c for c in df.select_dtypes('object').columns if c not in ['ID', 'location']]\n",
1739 |     "tmp = Parallel(n_jobs=4)(delayed(partial(pd.to_numeric, errors='coerce'))(df[c]) for c in obj_cols)\n",
1740 |     "df = df.drop(obj_cols, axis=1)\n",
1741 |     "df = pd.concat([df, pd.DataFrame(tmp).T], axis=1)\n",
1742 |     "weather_cols = ['temp', 'precip', 'rel_humidity', 'wind_dir','wind_spd', 'atmos_press']\n",
1743 |     "\n",
1744 |     "### Filling NaNs by interpolation.\n",
1745 |     "for w in tqdm_notebook(weather_cols):\n",
1746 |     "    selected_cols = [c for c in df.columns if w in c]\n",
1747 |     "    df[selected_cols] = df[selected_cols].interpolate(limit_direction='both')\n",
1748 |     "    ### Wind direction broken down into sine and cos components, and the original column is dropped\n",
1749 |     "    if w == 'wind_dir':\n",
1750 |     "        df[['wind_dir_cos_' + c.split('_')[-1] for c in selected_cols]] = df[selected_cols].apply(lambda x: np.cos(x))\n",
1751 |     "        df[['wind_dir_sin_' + c.split('_')[-1] for c in selected_cols]] = df[selected_cols].apply(lambda x: np.sin(x))\n",
1752 |     "        df = df.drop(selected_cols, axis = 1)    \n",
1753 |     "print(df.shape)\n",
1754 |     "features = [c for c in df.columns if c not in ['ID', 'location', 'target']]\n",
1755 |     "print(df[features].isnull().sum().sum())\n",
1756 |     "df.head()"
1757 |    ]
1758 |   },
1759 |   {
1760 |    "cell_type": "code",
1761 |    "execution_count": 9,
1762 |    "metadata": {
1763 |     "collapsed": true
1764 |    },
1765 |    "outputs": [],
1766 |    "source": [
1767 |     "### Encoding each location by the mean of its target\n",
1768 |     "features = [\"temp\",\"precip\",\"rel_humidity\",\"wind_spd\",\"atmos_press\", \"wind_dir_cos\", \"wind_dir_sin\"]\n",
1769 |     "df['location'] = df['location'].map(df.groupby('location')['target'].mean())"
1770 |    ]
1771 |   },
1772 |   {
1773 |    "cell_type": "code",
1774 |    "execution_count": 10,
1775 |    "metadata": {},
1776 |    "outputs": [
1777 |     {
1778 |      "name": "stderr",
1779 |      "output_type": "stream",
1780 |      "text": [
1781 |       "/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:4: TqdmDeprecationWarning: This function will be removed in tqdm==5.0.0\n",
1782 |       "Please use `tqdm.notebook.tqdm` instead of `tqdm.tqdm_notebook`\n",
1783 |       "  after removing the cwd from sys.path.\n"
1784 |      ]
1785 |     },
1786 |     {
1787 |      "data": {
1788 |       "application/vnd.jupyter.widget-view+json": {
1789 |        "model_id": "d3873ff95cce472896986a0bc35f665b",
1790 |        "version_major": 2,
1791 |        "version_minor": 0
1792 |       },
1793 |       "text/plain": [
1794 |        "HBox(children=(FloatProgress(value=0.0, max=7.0), HTML(value='')))"
1795 |       ]
1796 |      },
1797 |      "metadata": {},
1798 |      "output_type": "display_data"
1799 |     },
1800 |     {
1801 |      "name": "stdout",
1802 |      "output_type": "stream",
1803 |      "text": [
1804 |       "\n"
1805 |      ]
1806 |     }
1807 |    ],
1808 |    "source": [
1809 |     "### Creating more features\n",
1810 |     "\n",
1811 |     "new_features = []\n",
1812 |     "for f in tqdm_notebook(features):\n",
1813 |     "    new_features.append(f'{f}_24hrs_ratio')\n",
1814 |     "    new_features.append(f'{f}_6hrs_ratio')\n",
1815 |     "    for hour, f_col in enumerate([f\"f_{i}\" for i in range(121)]):\n",
1816 |     "        \n",
1817 |     "        ### Ratio of current hour and 24 hours before\n",
1818 |     "        if hour - 24 < 0:\n",
1819 |     "            df[f'{f}_24hrs_ratio_{hour}'] = df[f'{f}_{hour}'] /(1e-5 + df[f'{f}_0'])\n",
1820 |     "        else:\n",
1821 |     "            df[f'{f}_24hrs_ratio_{hour}'] = df[f'{f}_{hour}'] /(1e-5 + df[f'{f}_{hour-24}'])\n",
1822 |     "\n",
1823 |     "            \n",
1824 |     "        ### Ratio of current hour and 6 hours before\n",
1825 |     "        if hour - 6 < 0:\n",
1826 |     "            df[f'{f}_6hrs_ratio_{hour}'] = df[f'{f}_{hour}'] /(1e-5 + df[f'{f}_0'])\n",
1827 |     "        else:\n",
1828 |     "            df[f'{f}_6hrs_ratio_{hour}'] = df[f'{f}_{hour}'] /(1e-5 + df[f'{f}_{hour-6}'])"
1829 |    ]
1830 |   },
1831 |   {
1832 |    "cell_type": "code",
1833 |    "execution_count": 12,
1834 |    "metadata": {},
1835 |    "outputs": [
1836 |     {
1837 |      "data": {
1838 |       "text/plain": [
1839 |        "['temp',\n",
1840 |        " 'precip',\n",
1841 |        " 'rel_humidity',\n",
1842 |        " 'wind_spd',\n",
1843 |        " 'atmos_press',\n",
1844 |        " 'wind_dir_cos',\n",
1845 |        " 'wind_dir_sin',\n",
1846 |        " 'temp_24hrs_ratio',\n",
1847 |        " 'temp_6hrs_ratio',\n",
1848 |        " 'precip_24hrs_ratio',\n",
1849 |        " 'precip_6hrs_ratio',\n",
1850 |        " 'rel_humidity_24hrs_ratio',\n",
1851 |        " 'rel_humidity_6hrs_ratio',\n",
1852 |        " 'wind_spd_24hrs_ratio',\n",
1853 |        " 'wind_spd_6hrs_ratio',\n",
1854 |        " 'atmos_press_24hrs_ratio',\n",
1855 |        " 'atmos_press_6hrs_ratio',\n",
1856 |        " 'wind_dir_cos_24hrs_ratio',\n",
1857 |        " 'wind_dir_cos_6hrs_ratio',\n",
1858 |        " 'wind_dir_sin_24hrs_ratio',\n",
1859 |        " 'wind_dir_sin_6hrs_ratio']"
1860 |       ]
1861 |      },
1862 |      "execution_count": 12,
1863 |      "metadata": {},
1864 |      "output_type": "execute_result"
1865 |     }
1866 |    ],
1867 |    "source": [
1868 |     "features = features + new_features\n",
1869 |     "features"
1870 |    ]
1871 |   },
1872 |   {
1873 |    "cell_type": "code",
1874 |    "execution_count": 13,
1875 |    "metadata": {
1876 |     "collapsed": true
1877 |    },
1878 |    "outputs": [],
1879 |    "source": [
1880 |     "train=df[df.target.notnull()].reset_index(drop=True)\n",
1881 |     "test=df[df.target.isna()].reset_index(drop=True)"
1882 |    ]
1883 |   },
1884 |   {
1885 |    "cell_type": "code",
1886 |    "execution_count": 14,
1887 |    "metadata": {
1888 |     "collapsed": true
1889 |    },
1890 |    "outputs": [],
1891 |    "source": [
1892 |     "def get_sample(x):\n",
1893 |     "    '''\n",
1894 |     "    Convert a train row into a sample suitable for CNN\n",
1895 |     "    x: row of train dataset\n",
1896 |     "    '''\n",
1897 |     "    sub_sample = np.zeros((len(features)+1, 121))\n",
1898 |     "    for i, f in enumerate(features + [\"location\"]):\n",
1899 |     "        if f == \"location\":\n",
1900 |     "            sub_sample[i] = x[\"location\"] * np.ones(121)\n",
1901 |     "        else:\n",
1902 |     "            cols_f = [f\"{f}_{i}\" for i in range(121)]\n",
1903 |     "            vals = x[cols_f].values.astype(float)\n",
1904 |     "            sub_sample[i] = vals\n",
1905 |     "    return np.array(sub_sample).astype('float')"
1906 |    ]
1907 |   },
1908 |   {
1909 |    "cell_type": "code",
1910 |    "execution_count": 15,
1911 |    "metadata": {},
1912 |    "outputs": [
1913 |     {
1914 |      "data": {
1915 |       "text/plain": [
1916 |        "(15539, 2551)"
1917 |       ]
1918 |      },
1919 |      "execution_count": 15,
1920 |      "metadata": {},
1921 |      "output_type": "execute_result"
1922 |     }
1923 |    ],
1924 |    "source": [
1925 |     "train.shape"
1926 |    ]
1927 |   },
1928 |   {
1929 |    "cell_type": "code",
1930 |    "execution_count": 16,
1931 |    "metadata": {},
1932 |    "outputs": [
1933 |     {
1934 |      "name": "stdout",
1935 |      "output_type": "stream",
1936 |      "text": [
1937 |       "CPU times: user 1min 43s, sys: 3.45 s, total: 1min 47s\n",
1938 |       "Wall time: 6min 39s\n"
1939 |      ]
1940 |     }
1941 |    ],
1942 |    "source": [
1943 |     "%%time\n",
1944 |     "train_samples = Parallel(n_jobs=4)(delayed(get_sample)(row[1]) for row in train.iterrows())\n",
1945 |     "test_samples = Parallel(n_jobs=4)(delayed(get_sample)(row[1]) for row in test.iterrows())"
1946 |    ]
1947 |   },
1948 |   {
1949 |    "cell_type": "code",
1950 |    "execution_count": 17,
1951 |    "metadata": {},
1952 |    "outputs": [
1953 |     {
1954 |      "data": {
1955 |       "text/plain": [
1956 |        "((15539, 121, 22), (5035, 121, 22))"
1957 |       ]
1958 |      },
1959 |      "execution_count": 17,
1960 |      "metadata": {},
1961 |      "output_type": "execute_result"
1962 |     }
1963 |    ],
1964 |    "source": [
1965 |     "X_test = np.array(test_samples)\n",
1966 |     "X_test = np.einsum('ikj->ijk', X_test)\n",
1967 |     "X_train = np.array(train_samples)\n",
1968 |     "X_train = np.einsum('ikj->ijk', X_train)\n",
1969 |     "X_train.shape, X_test.shape"
1970 |    ]
1971 |   },
1972 |   {
1973 |    "cell_type": "code",
1974 |    "execution_count": 18,
1975 |    "metadata": {},
1976 |    "outputs": [
1977 |     {
1978 |      "name": "stdout",
1979 |      "output_type": "stream",
1980 |      "text": [
1981 |       "Train std =  1.020, Test std =  0.935, \n",
1982 |       "Train mean = -0.001, Test mean =  0.002\n"
1983 |      ]
1984 |     }
1985 |    ],
1986 |    "source": [
1987 |     "X_all = np.concatenate([X_train, X_test])\n",
1988 |     "\n",
1989 |     "my_mean = X_all.mean(axis=(0, 1))\n",
1990 |     "my_std = X_all.astype(float).std(axis=(0, 1))\n",
1991 |     "\n",
1992 |     "X_train[:, :] -= my_mean\n",
1993 |     "X_test [:, :] -= my_mean\n",
1994 |     "\n",
1995 |     "X_train[:, :] /= my_std\n",
1996 |     "X_test [:, :] /= my_std\n",
1997 |     "\n",
1998 |     "print(f\"Train std = {X_train.std(): .3f}, Test std = {X_test.std(): .3f}, \\nTrain mean = {X_train.mean(): .3f}, Test mean = {X_test.mean(): .3f}\")"
1999 |    ]
2000 |   },
2001 |   {
2002 |    "cell_type": "code",
2003 |    "execution_count": 19,
2004 |    "metadata": {},
2005 |    "outputs": [
2006 |     {
2007 |      "data": {
2008 |       "text/plain": [
2009 |        "(15539,)"
2010 |       ]
2011 |      },
2012 |      "execution_count": 19,
2013 |      "metadata": {},
2014 |      "output_type": "execute_result"
2015 |     }
2016 |    ],
2017 |    "source": [
2018 |     "y_train = train[\"target\"].values\n",
2019 |     "y_train.shape"
2020 |    ]
2021 |   },
2022 |   {
2023 |    "cell_type": "code",
2024 |    "execution_count": 20,
2025 |    "metadata": {},
2026 |    "outputs": [
2027 |     {
2028 |      "name": "stdout",
2029 |      "output_type": "stream",
2030 |      "text": [
2031 |       "Model: \"sequential\"\n",
2032 |       "_________________________________________________________________\n",
2033 |       "Layer (type)                 Output Shape              Param #   \n",
2034 |       "=================================================================\n",
2035 |       "batch_normalization (BatchNo multiple                  88        \n",
2036 |       "_________________________________________________________________\n",
2037 |       "conv1d (Conv1D)              multiple                  4288      \n",
2038 |       "_________________________________________________________________\n",
2039 |       "conv1d_1 (Conv1D)            multiple                  20544     \n",
2040 |       "_________________________________________________________________\n",
2041 |       "conv1d_2 (Conv1D)            multiple                  57472     \n",
2042 |       "_________________________________________________________________\n",
2043 |       "conv1d_3 (Conv1D)            multiple                  180352    \n",
2044 |       "_________________________________________________________________\n",
2045 |       "conv1d_4 (Conv1D)            multiple                  557312    \n",
2046 |       "_________________________________________________________________\n",
2047 |       "max_pooling1d (MaxPooling1D) multiple                  0         \n",
2048 |       "_________________________________________________________________\n",
2049 |       "flatten (Flatten)            multiple                  0         \n",
2050 |       "_________________________________________________________________\n",
2051 |       "batch_normalization_1 (Batch multiple                  20480     \n",
2052 |       "_________________________________________________________________\n",
2053 |       "dropout (Dropout)            multiple                  0         \n",
2054 |       "_________________________________________________________________\n",
2055 |       "output_cnn_simple (Dense)    multiple                  163872    \n",
2056 |       "_________________________________________________________________\n",
2057 |       "dense (Dense)                multiple                  33        \n",
2058 |       "=================================================================\n",
2059 |       "Total params: 1,004,441\n",
2060 |       "Trainable params: 994,157\n",
2061 |       "Non-trainable params: 10,284\n",
2062 |       "_________________________________________________________________\n"
2063 |      ]
2064 |     }
2065 |    ],
2066 |    "source": [
2067 |     "\n",
2068 |     "N_FEATS = X_train.shape[2]\n",
2069 |     "\n",
2070 |     "\n",
2071 |     "def root_mean_squared_error(y_true, y_pred):\n",
2072 |     "        return K.sqrt(K.mean(K.square(y_pred - y_true), axis=-1)) \n",
2073 |     "\n",
2074 |     "def make_model():\n",
2075 |     "    model = Sequential()\n",
2076 |     "    model.add(BatchNormalization())\n",
2077 |     "    model.add(Conv1D(filters=64, kernel_size=(3,), strides=1, activation='relu', input_shape=(121, N_FEATS)))\n",
2078 |     "    model.add(Conv1D(filters=64, kernel_size=(5 ,), strides=1, activation='relu'))\n",
2079 |     "    \n",
2080 |     "    model.add(Conv1D(filters=128, kernel_size=(7, ), strides=1, activation='relu'))\n",
2081 |     "    model.add(Conv1D(filters=128, kernel_size=(11, ), strides=1, activation='relu'))\n",
2082 |     "    model.add(Conv1D(filters=256, kernel_size=(17, ), strides=1, activation='relu'))\n",
2083 |     "    model.add(MaxPooling1D(4))\n",
2084 |     "    \n",
2085 |     "\n",
2086 |     "    model.add(Flatten())\n",
2087 |     "    model.add(BatchNormalization())\n",
2088 |     "    model.add(Dropout(0.3))\n",
2089 |     "    \n",
2090 |     "    model.add(Dense(32, activation='relu', name='output_cnn_simple'))\n",
2091 |     "    model.add(Dense(1))\n",
2092 |     "\n",
2093 |     "    model.compile(loss=root_mean_squared_error, optimizer=Adam(lr = 5e-4))\n",
2094 |     "    \n",
2095 |     "    return model\n",
2096 |     "\n",
2097 |     "m = make_model()\n",
2098 |     "m.build((None, 121, N_FEATS))\n",
2099 |     "m.summary()"
2100 |    ]
2101 |   },
2102 |   {
2103 |    "cell_type": "code",
2104 |    "execution_count": 21,
2105 |    "metadata": {},
2106 |    "outputs": [
2107 |     {
2108 |      "name": "stdout",
2109 |      "output_type": "stream",
2110 |      "text": [
2111 |       "\n",
2112 |       "_____________________Round n-1\n",
2113 |       "\n",
2114 |       "\n",
2115 |       "Fold n-1\n",
2116 |       "\n",
2117 |       "Restoring model weights from the end of the best epoch.\n",
2118 |       "Epoch 00060: early stopping\n",
2119 |       "Erreur = 23.69\n",
2120 |       "\n",
2121 |       "\n",
2122 |       "\n",
2123 |       "Fold n-2\n",
2124 |       "\n",
2125 |       "Restoring model weights from the end of the best epoch.\n",
2126 |       "Epoch 00050: early stopping\n",
2127 |       "Erreur = 23.92\n",
2128 |       "\n",
2129 |       "\n",
2130 |       "\n",
2131 |       "Fold n-3\n",
2132 |       "\n",
2133 |       "Restoring model weights from the end of the best epoch.\n",
2134 |       "Epoch 00069: early stopping\n",
2135 |       "Erreur = 20.91\n",
2136 |       "\n",
2137 |       "\n",
2138 |       "\n",
2139 |       "Fold n-4\n",
2140 |       "\n",
2141 |       "Restoring model weights from the end of the best epoch.\n",
2142 |       "Epoch 00039: early stopping\n",
2143 |       "Erreur = 22.98\n",
2144 |       "\n",
2145 |       "\n",
2146 |       "\n",
2147 |       "Fold n-5\n",
2148 |       "\n",
2149 |       "Restoring model weights from the end of the best epoch.\n",
2150 |       "Epoch 00055: early stopping\n",
2151 |       "Erreur = 21.24\n",
2152 |       "\n",
2153 |       "\n",
2154 |       "\n",
2155 |       "Fold n-6\n",
2156 |       "\n",
2157 |       "Restoring model weights from the end of the best epoch.\n",
2158 |       "Epoch 00046: early stopping\n",
2159 |       "Erreur = 24.60\n",
2160 |       "\n",
2161 |       "\n",
2162 |       "\n",
2163 |       "Fold n-7\n",
2164 |       "\n",
2165 |       "Restoring model weights from the end of the best epoch.\n",
2166 |       "Epoch 00054: early stopping\n",
2167 |       "Erreur = 24.30\n",
2168 |       "\n",
2169 |       "\n",
2170 |       "\n",
2171 |       "Fold n-8\n",
2172 |       "\n",
2173 |       "Restoring model weights from the end of the best epoch.\n",
2174 |       "Epoch 00045: early stopping\n",
2175 |       "Erreur = 24.49\n",
2176 |       "\n",
2177 |       "\n",
2178 |       "\n",
2179 |       "Fold n-9\n",
2180 |       "\n",
2181 |       "Restoring model weights from the end of the best epoch.\n",
2182 |       "Epoch 00046: early stopping\n",
2183 |       "Erreur = 23.19\n",
2184 |       "\n",
2185 |       "\n",
2186 |       "\n",
2187 |       "Fold n-10\n",
2188 |       "\n",
2189 |       "Restoring model weights from the end of the best epoch.\n",
2190 |       "Epoch 00050: early stopping\n",
2191 |       "Erreur = 22.19\n",
2192 |       "\n",
2193 |       "\n",
2194 |       "Total error =  23.151\n"
2195 |      ]
2196 |     }
2197 |    ],
2198 |    "source": [
2199 |     "errcb2 = list()\n",
2200 |     "y_pred_test = list()\n",
2201 |     "\n",
2202 |     "\n",
2203 |     "for my_round in range(1):\n",
2204 |     "    i = 0\n",
2205 |     "    print(f\"\\n_____________________Round n-{my_round + 1}\\n\")\n",
2206 |     "    \n",
2207 |     "    fold = KFold(n_splits=10, shuffle=True, random_state=my_round)\n",
2208 |     "    \n",
2209 |     "    for train_index, test_index in fold.split(X_train, y_train):\n",
2210 |     "        i += 1\n",
2211 |     "        print(f\"\\nFold n-{i}\\n\")\n",
2212 |     "\n",
2213 |     "        X_entr, X_val = X_train[train_index], X_train[test_index]\n",
2214 |     "        y_entr, y_val = y_train[train_index], y_train[test_index]\n",
2215 |     "\n",
2216 |     "        es = tf.keras.callbacks.EarlyStopping(patience=10, \n",
2217 |     "                                           verbose=1, \n",
2218 |     "                                           restore_best_weights=True)\n",
2219 |     "\n",
2220 |     "        m = make_model()\n",
2221 |     "\n",
2222 |     "        m.fit(X_entr,\n",
2223 |     "                y_entr,\n",
2224 |     "                epochs=150,\n",
2225 |     "                verbose=0,\n",
2226 |     "                batch_size=128,\n",
2227 |     "                validation_data=(X_val, y_val),\n",
2228 |     "                callbacks=[es],\n",
2229 |     "             )\n",
2230 |     "\n",
2231 |     "        # val score\n",
2232 |     "        preds = m.predict(X_val)[:, 0]\n",
2233 |     "        rmse_val = mean_squared_error(y_val, preds) ** .5\n",
2234 |     "        print(f\"Erreur = {rmse_val:.2f}\\n\\n\")\n",
2235 |     "        errcb2.append(rmse_val)\n",
2236 |     "\n",
2237 |     "        # prediction\n",
2238 |     "        p2 = m.predict(X_test)[:, 0]\n",
2239 |     "        p2 = np.clip(p2, train.target.min(), train.target.max())\n",
2240 |     "\n",
2241 |     "        y_pred_test.append(p2)\n",
2242 |     "\n",
2243 |     "print(f\"Total error = {np.mean(errcb2): .3f}\")"
2244 |    ]
2245 |   },
2246 |   {
2247 |    "cell_type": "code",
2248 |    "execution_count": 22,
2249 |    "metadata": {
2250 |     "collapsed": true
2251 |    },
2252 |    "outputs": [],
2253 |    "source": [
2254 |     "d = {'ID': test[\"ID\"], 'target': np.mean(y_pred_test, axis=0)}\n",
2255 |     "sub = pd.DataFrame(data=d)\n",
2256 |     "sub = sub[['ID', 'target']]\n",
2257 |     "sub['target'] = sub['target'].clip(train.target.min(), train.target.max())"
2258 |    ]
2259 |   },
2260 |   {
2261 |    "cell_type": "code",
2262 |    "execution_count": 25,
2263 |    "metadata": {},
2264 |    "outputs": [
2265 |     {
2266 |      "data": {
2267 |       "text/html": [
2268 |        "<div>\n",
2269 |        "<style scoped>\n",
2270 |        "    .dataframe tbody tr th:only-of-type {\n",
2271 |        "        vertical-align: middle;\n",
2272 |        "    }\n",
2273 |        "\n",
2274 |        "    .dataframe tbody tr th {\n",
2275 |        "        vertical-align: top;\n",
2276 |        "    }\n",
2277 |        "\n",
2278 |        "    .dataframe thead th {\n",
2279 |        "        text-align: right;\n",
2280 |        "    }\n",
2281 |        "</style>\n",
2282 |        "<table border=\"1\" class=\"dataframe\">\n",
2283 |        "  <thead>\n",
2284 |        "    <tr style=\"text-align: right;\">\n",
2285 |        "      <th></th>\n",
2286 |        "      <th>ID</th>\n",
2287 |        "      <th>target</th>\n",
2288 |        "    </tr>\n",
2289 |        "  </thead>\n",
2290 |        "  <tbody>\n",
2291 |        "    <tr>\n",
2292 |        "      <th>0</th>\n",
2293 |        "      <td>ID_test_0</td>\n",
2294 |        "      <td>146.448013</td>\n",
2295 |        "    </tr>\n",
2296 |        "    <tr>\n",
2297 |        "      <th>1</th>\n",
2298 |        "      <td>ID_test_1</td>\n",
2299 |        "      <td>78.831528</td>\n",
2300 |        "    </tr>\n",
2301 |        "    <tr>\n",
2302 |        "      <th>2</th>\n",
2303 |        "      <td>ID_test_10</td>\n",
2304 |        "      <td>33.348366</td>\n",
2305 |        "    </tr>\n",
2306 |        "    <tr>\n",
2307 |        "      <th>3</th>\n",
2308 |        "      <td>ID_test_100</td>\n",
2309 |        "      <td>55.782494</td>\n",
2310 |        "    </tr>\n",
2311 |        "    <tr>\n",
2312 |        "      <th>4</th>\n",
2313 |        "      <td>ID_test_1000</td>\n",
2314 |        "      <td>88.687622</td>\n",
2315 |        "    </tr>\n",
2316 |        "    <tr>\n",
2317 |        "      <th>5</th>\n",
2318 |        "      <td>ID_test_1001</td>\n",
2319 |        "      <td>30.877085</td>\n",
2320 |        "    </tr>\n",
2321 |        "    <tr>\n",
2322 |        "      <th>6</th>\n",
2323 |        "      <td>ID_test_1002</td>\n",
2324 |        "      <td>75.518860</td>\n",
2325 |        "    </tr>\n",
2326 |        "    <tr>\n",
2327 |        "      <th>7</th>\n",
2328 |        "      <td>ID_test_1003</td>\n",
2329 |        "      <td>37.986477</td>\n",
2330 |        "    </tr>\n",
2331 |        "    <tr>\n",
2332 |        "      <th>8</th>\n",
2333 |        "      <td>ID_test_1004</td>\n",
2334 |        "      <td>34.908413</td>\n",
2335 |        "    </tr>\n",
2336 |        "    <tr>\n",
2337 |        "      <th>9</th>\n",
2338 |        "      <td>ID_test_1005</td>\n",
2339 |        "      <td>52.110146</td>\n",
2340 |        "    </tr>\n",
2341 |        "  </tbody>\n",
2342 |        "</table>\n",
2343 |        "</div>"
2344 |       ],
2345 |       "text/plain": [
2346 |        "             ID      target\n",
2347 |        "0     ID_test_0  146.448013\n",
2348 |        "1     ID_test_1   78.831528\n",
2349 |        "2    ID_test_10   33.348366\n",
2350 |        "3   ID_test_100   55.782494\n",
2351 |        "4  ID_test_1000   88.687622\n",
2352 |        "5  ID_test_1001   30.877085\n",
2353 |        "6  ID_test_1002   75.518860\n",
2354 |        "7  ID_test_1003   37.986477\n",
2355 |        "8  ID_test_1004   34.908413\n",
2356 |        "9  ID_test_1005   52.110146"
2357 |       ]
2358 |      },
2359 |      "execution_count": 25,
2360 |      "metadata": {},
2361 |      "output_type": "execute_result"
2362 |     }
2363 |    ],
2364 |    "source": [
2365 |     "sub.head(10)"
2366 |    ]
2367 |   },
2368 |   {
2369 |    "cell_type": "code",
2370 |    "execution_count": 29,
2371 |    "metadata": {},
2372 |    "outputs": [
2373 |     {
2374 |      "data": {
2375 |       "text/plain": [
2376 |        "23.151077529050763"
2377 |       ]
2378 |      },
2379 |      "execution_count": 29,
2380 |      "metadata": {},
2381 |      "output_type": "execute_result"
2382 |     }
2383 |    ],
2384 |    "source": [
2385 |     "np.mean(errcb2)"
2386 |    ]
2387 |   },
2388 |   {
2389 |    "cell_type": "code",
2390 |    "execution_count": 32,
2391 |    "metadata": {
2392 |     "collapsed": true
2393 |    },
2394 |    "outputs": [],
2395 |    "source": [
2396 |     "sub.to_csv(f\"cnn_preds.csv\", index=False)"
2397 |    ]
2398 |   },
2399 |   {
2400 |    "cell_type": "code",
2401 |    "execution_count": 33,
2402 |    "metadata": {},
2403 |    "outputs": [
2404 |     {
2405 |      "data": {
2406 |       "text/html": [
2407 |        "<a href=cnn_preds.csv>Download CSV file</a>"
2408 |       ],
2409 |       "text/plain": [
2410 |        "<IPython.core.display.HTML object>"
2411 |       ]
2412 |      },
2413 |      "execution_count": 33,
2414 |      "metadata": {},
2415 |      "output_type": "execute_result"
2416 |     }
2417 |    ],
2418 |    "source": [
2419 |     "SUB_FILE_NAME = f\"cnn_preds.csv\"\n",
2420 |     "from IPython.display import HTML\n",
2421 |     "def create_download_link(title = \"Download CSV file\", filename = \"data.csv\"):  \n",
2422 |     "    html = '<a href={filename}>{title}</a>'\n",
2423 |     "    html = html.format(title=title,filename=filename)\n",
2424 |     "    return HTML(html)\n",
2425 |     "create_download_link(filename = SUB_FILE_NAME)"
2426 |    ]
2427 |   }
2428 |  ],
2429 |  "metadata": {
2430 |   "kernelspec": {
2431 |    "display_name": "Python 3",
2432 |    "language": "python",
2433 |    "name": "python3"
2434 |   },
2435 |   "language_info": {
2436 |    "codemirror_mode": {
2437 |     "name": "ipython",
2438 |     "version": 3
2439 |    },
2440 |    "file_extension": ".py",
2441 |    "mimetype": "text/x-python",
2442 |    "name": "python",
2443 |    "nbconvert_exporter": "python",
2444 |    "pygments_lexer": "ipython3",
2445 |    "version": "3.6.3"
2446 |   }
2447 |  },
2448 |  "nbformat": 4,
2449 |  "nbformat_minor": 4
2450 | }
2451 | 


--------------------------------------------------------------------------------