├── README.md
├── meli2021
    ├── 19_xgb_barely_tuned_yt.ipynb
    ├── 32_baseline_yt.ipynb
    ├── 61_active_model_yt.ipynb
    ├── README
    └── utils.py
└── multiple_time_series
    ├── README
    └── workshop_notebook.ipynb


/README.md:
--------------------------------------------------------------------------------
1 | # english_tutorials


--------------------------------------------------------------------------------
/meli2021/19_xgb_barely_tuned_yt.ipynb:
--------------------------------------------------------------------------------
   1 | {
   2 |  "cells": [
   3 |   {
   4 |    "cell_type": "code",
   5 |    "execution_count": 1,
   6 |    "id": "75638528-4609-4ee2-93c9-28538579e471",
   7 |    "metadata": {},
   8 |    "outputs": [],
   9 |    "source": [
  10 |     "import pandas as pd\n",
  11 |     "import numpy as np\n",
  12 |     "import utils\n",
  13 |     "\n",
  14 |     "from sklearn.model_selection import GroupKFold, KFold\n",
  15 |     "from sklearn.linear_model import LinearRegression\n",
  16 |     "from sklearn.ensemble import RandomForestRegressor\n",
  17 |     "from sklearn.metrics import mean_squared_error\n",
  18 |     "from xgboost import XGBRegressor\n",
  19 |     "import tweedie\n",
  20 |     "\n",
  21 |     "\n",
  22 |     "from importlib import reload\n",
  23 |     "reload(utils)\n",
  24 |     "from skopt import gp_minimize"
  25 |    ]
  26 |   },
  27 |   {
  28 |    "cell_type": "code",
  29 |    "execution_count": 2,
  30 |    "id": "3be069d6-c17e-43a2-8252-2acf5e9c26ad",
  31 |    "metadata": {},
  32 |    "outputs": [],
  33 |    "source": [
  34 |     "train = pd.read_parquet(\"./train/0.parquet\")\n",
  35 |     "train['date'] = pd.to_datetime(train['date'])\n",
  36 |     "train['fold'] = train['date'].dt.month"
  37 |    ]
  38 |   },
  39 |   {
  40 |    "cell_type": "code",
  41 |    "execution_count": 3,
  42 |    "id": "b9975ad7-ff3a-4b58-9d4c-b1435d0c5535",
  43 |    "metadata": {},
  44 |    "outputs": [
  45 |     {
  46 |      "data": {
  47 |       "text/html": [
  48 |        "<div>\n",
  49 |        "<style scoped>\n",
  50 |        "    .dataframe tbody tr th:only-of-type {\n",
  51 |        "        vertical-align: middle;\n",
  52 |        "    }\n",
  53 |        "\n",
  54 |        "    .dataframe tbody tr th {\n",
  55 |        "        vertical-align: top;\n",
  56 |        "    }\n",
  57 |        "\n",
  58 |        "    .dataframe thead th {\n",
  59 |        "        text-align: right;\n",
  60 |        "    }\n",
  61 |        "</style>\n",
  62 |        "<table border=\"1\" class=\"dataframe\">\n",
  63 |        "  <thead>\n",
  64 |        "    <tr style=\"text-align: right;\">\n",
  65 |        "      <th></th>\n",
  66 |        "      <th>sku</th>\n",
  67 |        "      <th>date</th>\n",
  68 |        "      <th>sold_quantity</th>\n",
  69 |        "      <th>current_price</th>\n",
  70 |        "      <th>currency</th>\n",
  71 |        "      <th>listing_type</th>\n",
  72 |        "      <th>shipping_logistic_type</th>\n",
  73 |        "      <th>shipping_payment</th>\n",
  74 |        "      <th>minutes_active</th>\n",
  75 |        "      <th>item_domain_id</th>\n",
  76 |        "      <th>site_id</th>\n",
  77 |        "      <th>fold</th>\n",
  78 |        "    </tr>\n",
  79 |        "  </thead>\n",
  80 |        "  <tbody>\n",
  81 |        "    <tr>\n",
  82 |        "      <th>0</th>\n",
  83 |        "      <td>464801</td>\n",
  84 |        "      <td>2021-02-01</td>\n",
  85 |        "      <td>0</td>\n",
  86 |        "      <td>156.78</td>\n",
  87 |        "      <td>REA</td>\n",
  88 |        "      <td>classic</td>\n",
  89 |        "      <td>fulfillment</td>\n",
  90 |        "      <td>free_shipping</td>\n",
  91 |        "      <td>1440.0</td>\n",
  92 |        "      <td>MLB-NEBULIZERS</td>\n",
  93 |        "      <td>MLB</td>\n",
  94 |        "      <td>2</td>\n",
  95 |        "    </tr>\n",
  96 |        "    <tr>\n",
  97 |        "      <th>1</th>\n",
  98 |        "      <td>464801</td>\n",
  99 |        "      <td>2021-02-02</td>\n",
 100 |        "      <td>0</td>\n",
 101 |        "      <td>156.78</td>\n",
 102 |        "      <td>REA</td>\n",
 103 |        "      <td>classic</td>\n",
 104 |        "      <td>fulfillment</td>\n",
 105 |        "      <td>free_shipping</td>\n",
 106 |        "      <td>1440.0</td>\n",
 107 |        "      <td>MLB-NEBULIZERS</td>\n",
 108 |        "      <td>MLB</td>\n",
 109 |        "      <td>2</td>\n",
 110 |        "    </tr>\n",
 111 |        "    <tr>\n",
 112 |        "      <th>2</th>\n",
 113 |        "      <td>464801</td>\n",
 114 |        "      <td>2021-02-03</td>\n",
 115 |        "      <td>0</td>\n",
 116 |        "      <td>156.78</td>\n",
 117 |        "      <td>REA</td>\n",
 118 |        "      <td>classic</td>\n",
 119 |        "      <td>fulfillment</td>\n",
 120 |        "      <td>free_shipping</td>\n",
 121 |        "      <td>1440.0</td>\n",
 122 |        "      <td>MLB-NEBULIZERS</td>\n",
 123 |        "      <td>MLB</td>\n",
 124 |        "      <td>2</td>\n",
 125 |        "    </tr>\n",
 126 |        "    <tr>\n",
 127 |        "      <th>3</th>\n",
 128 |        "      <td>464801</td>\n",
 129 |        "      <td>2021-02-04</td>\n",
 130 |        "      <td>0</td>\n",
 131 |        "      <td>156.78</td>\n",
 132 |        "      <td>REA</td>\n",
 133 |        "      <td>classic</td>\n",
 134 |        "      <td>fulfillment</td>\n",
 135 |        "      <td>free_shipping</td>\n",
 136 |        "      <td>1440.0</td>\n",
 137 |        "      <td>MLB-NEBULIZERS</td>\n",
 138 |        "      <td>MLB</td>\n",
 139 |        "      <td>2</td>\n",
 140 |        "    </tr>\n",
 141 |        "    <tr>\n",
 142 |        "      <th>4</th>\n",
 143 |        "      <td>464801</td>\n",
 144 |        "      <td>2021-02-05</td>\n",
 145 |        "      <td>1</td>\n",
 146 |        "      <td>156.78</td>\n",
 147 |        "      <td>REA</td>\n",
 148 |        "      <td>classic</td>\n",
 149 |        "      <td>fulfillment</td>\n",
 150 |        "      <td>free_shipping</td>\n",
 151 |        "      <td>1440.0</td>\n",
 152 |        "      <td>MLB-NEBULIZERS</td>\n",
 153 |        "      <td>MLB</td>\n",
 154 |        "      <td>2</td>\n",
 155 |        "    </tr>\n",
 156 |        "  </tbody>\n",
 157 |        "</table>\n",
 158 |        "</div>"
 159 |       ],
 160 |       "text/plain": [
 161 |        "      sku       date  sold_quantity  current_price currency listing_type  \\\n",
 162 |        "0  464801 2021-02-01              0         156.78      REA      classic   \n",
 163 |        "1  464801 2021-02-02              0         156.78      REA      classic   \n",
 164 |        "2  464801 2021-02-03              0         156.78      REA      classic   \n",
 165 |        "3  464801 2021-02-04              0         156.78      REA      classic   \n",
 166 |        "4  464801 2021-02-05              1         156.78      REA      classic   \n",
 167 |        "\n",
 168 |        "  shipping_logistic_type shipping_payment  minutes_active  item_domain_id  \\\n",
 169 |        "0            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
 170 |        "1            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
 171 |        "2            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
 172 |        "3            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
 173 |        "4            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
 174 |        "\n",
 175 |        "  site_id  fold  \n",
 176 |        "0     MLB     2  \n",
 177 |        "1     MLB     2  \n",
 178 |        "2     MLB     2  \n",
 179 |        "3     MLB     2  \n",
 180 |        "4     MLB     2  "
 181 |       ]
 182 |      },
 183 |      "execution_count": 3,
 184 |      "metadata": {},
 185 |      "output_type": "execute_result"
 186 |     }
 187 |    ],
 188 |    "source": [
 189 |     "train.head()"
 190 |    ]
 191 |   },
 192 |   {
 193 |    "cell_type": "code",
 194 |    "execution_count": 3,
 195 |    "id": "smaller-boulder",
 196 |    "metadata": {},
 197 |    "outputs": [],
 198 |    "source": [
 199 |     "test = pd.read_csv(\"test_data.csv\", index_col=0).squeeze()"
 200 |    ]
 201 |   },
 202 |   {
 203 |    "cell_type": "code",
 204 |    "execution_count": 4,
 205 |    "id": "finnish-canadian",
 206 |    "metadata": {},
 207 |    "outputs": [
 208 |     {
 209 |      "data": {
 210 |       "text/plain": [
 211 |        "sku                                int64\n",
 212 |        "date                      datetime64[ns]\n",
 213 |        "sold_quantity                      int64\n",
 214 |        "current_price                    float64\n",
 215 |        "currency                          object\n",
 216 |        "listing_type                      object\n",
 217 |        "shipping_logistic_type            object\n",
 218 |        "shipping_payment                  object\n",
 219 |        "minutes_active                   float64\n",
 220 |        "item_domain_id                    object\n",
 221 |        "site_id                           object\n",
 222 |        "fold                               int64\n",
 223 |        "dtype: object"
 224 |       ]
 225 |      },
 226 |      "execution_count": 4,
 227 |      "metadata": {},
 228 |      "output_type": "execute_result"
 229 |     }
 230 |    ],
 231 |    "source": [
 232 |     "train.dtypes"
 233 |    ]
 234 |   },
 235 |   {
 236 |    "cell_type": "code",
 237 |    "execution_count": 5,
 238 |    "id": "fc10499b-4c46-4f7e-a88e-a4233fb05504",
 239 |    "metadata": {},
 240 |    "outputs": [
 241 |     {
 242 |      "data": {
 243 |       "text/html": [
 244 |        "<div>\n",
 245 |        "<style scoped>\n",
 246 |        "    .dataframe tbody tr th:only-of-type {\n",
 247 |        "        vertical-align: middle;\n",
 248 |        "    }\n",
 249 |        "\n",
 250 |        "    .dataframe tbody tr th {\n",
 251 |        "        vertical-align: top;\n",
 252 |        "    }\n",
 253 |        "\n",
 254 |        "    .dataframe thead th {\n",
 255 |        "        text-align: right;\n",
 256 |        "    }\n",
 257 |        "</style>\n",
 258 |        "<table border=\"1\" class=\"dataframe\">\n",
 259 |        "  <thead>\n",
 260 |        "    <tr style=\"text-align: right;\">\n",
 261 |        "      <th></th>\n",
 262 |        "      <th>sku</th>\n",
 263 |        "      <th>date</th>\n",
 264 |        "      <th>sold_quantity</th>\n",
 265 |        "      <th>current_price</th>\n",
 266 |        "      <th>currency</th>\n",
 267 |        "      <th>listing_type</th>\n",
 268 |        "      <th>shipping_logistic_type</th>\n",
 269 |        "      <th>shipping_payment</th>\n",
 270 |        "      <th>minutes_active</th>\n",
 271 |        "      <th>item_domain_id</th>\n",
 272 |        "      <th>site_id</th>\n",
 273 |        "      <th>fold</th>\n",
 274 |        "    </tr>\n",
 275 |        "  </thead>\n",
 276 |        "  <tbody>\n",
 277 |        "    <tr>\n",
 278 |        "      <th>0</th>\n",
 279 |        "      <td>464801</td>\n",
 280 |        "      <td>2021-02-01</td>\n",
 281 |        "      <td>0</td>\n",
 282 |        "      <td>156.78</td>\n",
 283 |        "      <td>REA</td>\n",
 284 |        "      <td>classic</td>\n",
 285 |        "      <td>fulfillment</td>\n",
 286 |        "      <td>free_shipping</td>\n",
 287 |        "      <td>1440.0</td>\n",
 288 |        "      <td>MLB-NEBULIZERS</td>\n",
 289 |        "      <td>MLB</td>\n",
 290 |        "      <td>2</td>\n",
 291 |        "    </tr>\n",
 292 |        "    <tr>\n",
 293 |        "      <th>1</th>\n",
 294 |        "      <td>464801</td>\n",
 295 |        "      <td>2021-02-02</td>\n",
 296 |        "      <td>0</td>\n",
 297 |        "      <td>156.78</td>\n",
 298 |        "      <td>REA</td>\n",
 299 |        "      <td>classic</td>\n",
 300 |        "      <td>fulfillment</td>\n",
 301 |        "      <td>free_shipping</td>\n",
 302 |        "      <td>1440.0</td>\n",
 303 |        "      <td>MLB-NEBULIZERS</td>\n",
 304 |        "      <td>MLB</td>\n",
 305 |        "      <td>2</td>\n",
 306 |        "    </tr>\n",
 307 |        "    <tr>\n",
 308 |        "      <th>2</th>\n",
 309 |        "      <td>464801</td>\n",
 310 |        "      <td>2021-02-03</td>\n",
 311 |        "      <td>0</td>\n",
 312 |        "      <td>156.78</td>\n",
 313 |        "      <td>REA</td>\n",
 314 |        "      <td>classic</td>\n",
 315 |        "      <td>fulfillment</td>\n",
 316 |        "      <td>free_shipping</td>\n",
 317 |        "      <td>1440.0</td>\n",
 318 |        "      <td>MLB-NEBULIZERS</td>\n",
 319 |        "      <td>MLB</td>\n",
 320 |        "      <td>2</td>\n",
 321 |        "    </tr>\n",
 322 |        "    <tr>\n",
 323 |        "      <th>3</th>\n",
 324 |        "      <td>464801</td>\n",
 325 |        "      <td>2021-02-04</td>\n",
 326 |        "      <td>0</td>\n",
 327 |        "      <td>156.78</td>\n",
 328 |        "      <td>REA</td>\n",
 329 |        "      <td>classic</td>\n",
 330 |        "      <td>fulfillment</td>\n",
 331 |        "      <td>free_shipping</td>\n",
 332 |        "      <td>1440.0</td>\n",
 333 |        "      <td>MLB-NEBULIZERS</td>\n",
 334 |        "      <td>MLB</td>\n",
 335 |        "      <td>2</td>\n",
 336 |        "    </tr>\n",
 337 |        "    <tr>\n",
 338 |        "      <th>4</th>\n",
 339 |        "      <td>464801</td>\n",
 340 |        "      <td>2021-02-05</td>\n",
 341 |        "      <td>1</td>\n",
 342 |        "      <td>156.78</td>\n",
 343 |        "      <td>REA</td>\n",
 344 |        "      <td>classic</td>\n",
 345 |        "      <td>fulfillment</td>\n",
 346 |        "      <td>free_shipping</td>\n",
 347 |        "      <td>1440.0</td>\n",
 348 |        "      <td>MLB-NEBULIZERS</td>\n",
 349 |        "      <td>MLB</td>\n",
 350 |        "      <td>2</td>\n",
 351 |        "    </tr>\n",
 352 |        "  </tbody>\n",
 353 |        "</table>\n",
 354 |        "</div>"
 355 |       ],
 356 |       "text/plain": [
 357 |        "      sku       date  sold_quantity  current_price currency listing_type  \\\n",
 358 |        "0  464801 2021-02-01              0         156.78      REA      classic   \n",
 359 |        "1  464801 2021-02-02              0         156.78      REA      classic   \n",
 360 |        "2  464801 2021-02-03              0         156.78      REA      classic   \n",
 361 |        "3  464801 2021-02-04              0         156.78      REA      classic   \n",
 362 |        "4  464801 2021-02-05              1         156.78      REA      classic   \n",
 363 |        "\n",
 364 |        "  shipping_logistic_type shipping_payment  minutes_active  item_domain_id  \\\n",
 365 |        "0            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
 366 |        "1            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
 367 |        "2            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
 368 |        "3            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
 369 |        "4            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
 370 |        "\n",
 371 |        "  site_id  fold  \n",
 372 |        "0     MLB     2  \n",
 373 |        "1     MLB     2  \n",
 374 |        "2     MLB     2  \n",
 375 |        "3     MLB     2  \n",
 376 |        "4     MLB     2  "
 377 |       ]
 378 |      },
 379 |      "execution_count": 5,
 380 |      "metadata": {},
 381 |      "output_type": "execute_result"
 382 |     }
 383 |    ],
 384 |    "source": [
 385 |     "train.head()"
 386 |    ]
 387 |   },
 388 |   {
 389 |    "cell_type": "code",
 390 |    "execution_count": 6,
 391 |    "id": "refined-string",
 392 |    "metadata": {},
 393 |    "outputs": [],
 394 |    "source": [
 395 |     "cats = ['item_domain_id', 'currency', 'listing_type', 'shipping_logistic_type', 'shipping_payment', 'site_id']"
 396 |    ]
 397 |   },
 398 |   {
 399 |    "cell_type": "code",
 400 |    "execution_count": 7,
 401 |    "id": "99b8015a-8c95-4936-9ba6-9d655aa19848",
 402 |    "metadata": {},
 403 |    "outputs": [],
 404 |    "source": [
 405 |     "from category_encoders import OrdinalEncoder\n",
 406 |     "enc = OrdinalEncoder(cats)\n",
 407 |     "train = enc.fit_transform(train)"
 408 |    ]
 409 |   },
 410 |   {
 411 |    "cell_type": "code",
 412 |    "execution_count": 8,
 413 |    "id": "monthly-general",
 414 |    "metadata": {},
 415 |    "outputs": [],
 416 |    "source": [
 417 |     "def gen_tr_ts():\n",
 418 |     "    for fold in [2,3]:\n",
 419 |     "        ts = train[train['fold'] != fold]['date'].max()\n",
 420 |     "        ts = train[(train['fold'] != fold) & (train['date'] == ts)].index\n",
 421 |     "        yield train.index[train['fold'] == fold], ts, fold\n",
 422 |     "\n",
 423 |     "    "
 424 |    ]
 425 |   },
 426 |   {
 427 |    "cell_type": "code",
 428 |    "execution_count": null,
 429 |    "id": "bdb25520-ed36-4c4c-a9fe-23387cb3f918",
 430 |    "metadata": {},
 431 |    "outputs": [
 432 |     {
 433 |      "name": "stdout",
 434 |      "output_type": "stream",
 435 |      "text": [
 436 |       "Iteration No: 1 started. Evaluating function at random point.\n",
 437 |       "[0.09871192514273254, 9, 0.16531200313642108, 0.9491364637917304, 1.2337280871824563, 120]\n",
 438 |       "8.619976237016095\n",
 439 |       "8.863340114023996\n",
 440 |       "Iteration No: 1 ended. Evaluation done at random point.\n",
 441 |       "Time taken: 164.3399\n",
 442 |       "Function value obtained: 8.7417\n",
 443 |       "Current minimum: 8.7417\n",
 444 |       "Iteration No: 2 started. Evaluating function at random point.\n",
 445 |       "[0.0059678992438367785, 7, 0.8919851637254288, 0.8116798250174155, 1.3101407817629525, 158]\n",
 446 |       "8.969453535454983\n",
 447 |       "9.180772268605745\n",
 448 |       "Iteration No: 2 ended. Evaluation done at random point.\n",
 449 |       "Time taken: 161.5522\n",
 450 |       "Function value obtained: 9.0751\n",
 451 |       "Current minimum: 8.7417\n",
 452 |       "Iteration No: 3 started. Evaluating function at random point.\n",
 453 |       "[0.007707362534461022, 3, 0.5309725180523154, 0.8725658221213098, 1.4526327599071185, 130]\n",
 454 |       "9.148086951536671\n"
 455 |      ]
 456 |     }
 457 |    ],
 458 |    "source": [
 459 |     "def tune(params):\n",
 460 |     "    print(params)\n",
 461 |     "    features = [\"current_price\", \"minutes_active\"] + cats\n",
 462 |     "\n",
 463 |     "    mean_rps = 0.\n",
 464 |     "    for tr,ts, fold in gen_tr_ts():\n",
 465 |     "        #print(tr.shape, ts.shape)\n",
 466 |     "        X = train[features]\n",
 467 |     "        y = train['sold_quantity']\n",
 468 |     "\n",
 469 |     "        Xtr = X.iloc[tr]\n",
 470 |     "        ytr = y.iloc[tr]\n",
 471 |     "        Xval = X.iloc[ts]\n",
 472 |     "        yval = y.iloc[ts]\n",
 473 |     "\n",
 474 |     "        #mdl = LinearRegression(normalize=True)\n",
 475 |     "        #mdl = RandomForestRegressor(n_estimators=100, random_state=0, n_jobs=6)\n",
 476 |     "        mdl = XGBRegressor(n_estimators=1000, learning_rate=params[0],\n",
 477 |     "                           max_depth=params[1],\n",
 478 |     "                           subsample=params[2],\n",
 479 |     "                           colsample_bytree=params[3],\n",
 480 |     "                           tweedie_variance_power=params[4],\n",
 481 |     "                           min_child_weight=params[5],\n",
 482 |     "                           random_state=0, objective=\"reg:tweedie\", \n",
 483 |     "                           base_score=1e-3,\n",
 484 |     "                           tree_method='gpu_hist')\n",
 485 |     "        mdl.fit(Xtr, ytr)\n",
 486 |     "        p = mdl.predict(Xval)\n",
 487 |     "\n",
 488 |     "\n",
 489 |     "        ## EVAL\n",
 490 |     "        pp = train[train['fold'] != fold][['sku', 'date', 'sold_quantity']]\n",
 491 |     "        pp['stock'] = pp['sku'].map(test)\n",
 492 |     "        pp = pp.sort_values([\"sku\",\"date\"])\n",
 493 |     "        pp['cumulative_y'] = pp.groupby(\"sku\")['sold_quantity'].cumsum()\n",
 494 |     "\n",
 495 |     "        pp = pp.dropna(subset=['stock'])\n",
 496 |     "        pp['stockout_y'] = pp['cumulative_y'] >= pp['stock']\n",
 497 |     "\n",
 498 |     "        first_so_y = pp[pp['stockout_y']].groupby(\"sku\").first()\n",
 499 |     "        days_to_so_y = (first_so_y[\"date\"] - pp[\"date\"].min()) / np.timedelta64(1, 'D')\n",
 500 |     "        days_to_so_y = days_to_so_y.reindex(pp['sku'].unique()).fillna(30.).clip(1,30)\n",
 501 |     "\n",
 502 |     "\n",
 503 |     "        ppp = train.iloc[ts][['sku']]\n",
 504 |     "        #p[~np.isfinite(p)] = 17.\n",
 505 |     "        ppp['p'] = p\n",
 506 |     "        ppp['stock'] = ppp['sku'].map(test)\n",
 507 |     "        ppp = ppp.dropna(subset=['stock'])\n",
 508 |     "        ppp['days_to_so'] = (ppp['stock'] / ppp['p']).astype(int).fillna(30.).clip(1,30)\n",
 509 |     "        days_to_so_p = ppp[['sku', 'days_to_so']].set_index(\"sku\").squeeze().reindex(days_to_so_y.index) \n",
 510 |     "\n",
 511 |     "        days_to_so_p2 = utils.pred_list_to_tweedie(days_to_so_p, phi=2, p=1.5)\n",
 512 |     "        \n",
 513 |     "        #tweedie distribution -> [0.05, 0.07, ... .13, 0.12]\n",
 514 |     "\n",
 515 |     "        rps = utils.rps(days_to_so_y, days_to_so_p2, probs=True)\n",
 516 |     "        mean_rps += rps\n",
 517 |     "        print(rps)\n",
 518 |     "    return mean_rps / 2\n",
 519 |     "\n",
 520 |     "space = [(1e-3, 1e-1, 'log-uniform'),\n",
 521 |     "         (1, 10),\n",
 522 |     "         (0.05, 0.95),\n",
 523 |     "         (0.05, 0.95),\n",
 524 |     "         (1.0,1.99),\n",
 525 |     "         (1,300)]\n",
 526 |     "res = gp_minimize(tune, space, random_state=1, verbose=1)\n",
 527 |     "\n",
 528 |     "    "
 529 |    ]
 530 |   },
 531 |   {
 532 |    "cell_type": "code",
 533 |    "execution_count": null,
 534 |    "id": "pretty-literature",
 535 |    "metadata": {},
 536 |    "outputs": [],
 537 |    "source": [
 538 |     "# 15a \n",
 539 |     "# Mean CV 9.0805\n",
 540 |     "# LB 6.2598\n",
 541 |     "\n",
 542 |     "\n",
 543 |     "Iteration No: 2 started. Evaluating function at random point.\n",
 544 |     "[0.003936128001463711, 2, 0.29539066512210194, 0.47989860558921493, 1.8040470414877383, 145]\n",
 545 |     "6.131413939395725\n",
 546 |     "6.4664243315180086\n",
 547 |     "Iteration No: 2 ended. Evaluation done at random point.\n",
 548 |     "Time taken: 91.9157\n",
 549 |     "Function value obtained: 6.2989\n",
 550 |     "Current minimum: 6.2989"
 551 |    ]
 552 |   },
 553 |   {
 554 |    "cell_type": "markdown",
 555 |    "id": "comfortable-hypothesis",
 556 |    "metadata": {},
 557 |    "source": [
 558 |     "# sub"
 559 |    ]
 560 |   },
 561 |   {
 562 |    "cell_type": "code",
 563 |    "execution_count": 10,
 564 |    "id": "dominant-machine",
 565 |    "metadata": {},
 566 |    "outputs": [
 567 |     {
 568 |      "name": "stdout",
 569 |      "output_type": "stream",
 570 |      "text": [
 571 |       "True\n"
 572 |      ]
 573 |     }
 574 |    ],
 575 |    "source": [
 576 |     "test_df = train[train['date'] == \"2021-03-31\"]\n",
 577 |     "test_df = test_df[test_df['sku'].isin(test.index)]\n",
 578 |     "print(np.all(test_df['sku'] == test.index))\n",
 579 |     "\n",
 580 |     "features = [\"current_price\", \"minutes_active\"] + cats\n",
 581 |     "params = [0.003936128001463711, 2, 0.29539066512210194, 0.47989860558921493, 1.8040470414877383, 145]\n",
 582 |     "mdl = XGBRegressor(n_estimators=1000, learning_rate=params[0],\n",
 583 |     "                   max_depth=params[1],\n",
 584 |     "                   subsample=params[2],\n",
 585 |     "                   colsample_bytree=params[3],\n",
 586 |     "                   tweedie_variance_power=params[4],\n",
 587 |     "                   min_child_weight=params[5],\n",
 588 |     "                   random_state=0, objective=\"reg:tweedie\", \n",
 589 |     "                   base_score=1e-3,\n",
 590 |     "                   tree_method='gpu_hist')\n",
 591 |     "mdl.fit(train[features], train['sold_quantity'])\n",
 592 |     "p = mdl.predict(test_df[features])"
 593 |    ]
 594 |   },
 595 |   {
 596 |    "cell_type": "code",
 597 |    "execution_count": 14,
 598 |    "id": "important-rugby",
 599 |    "metadata": {},
 600 |    "outputs": [],
 601 |    "source": [
 602 |     "spp = test_df[['sku']].copy()\n",
 603 |     "spp['p'] = p\n",
 604 |     "spp['stock'] = spp['sku'].map(test)\n",
 605 |     "spp['days_to_so'] = (spp['stock'] / spp['p']).fillna(30.).clip(1,30).astype(int)\n"
 606 |    ]
 607 |   },
 608 |   {
 609 |    "cell_type": "code",
 610 |    "execution_count": 15,
 611 |    "id": "excess-porter",
 612 |    "metadata": {},
 613 |    "outputs": [
 614 |     {
 615 |      "data": {
 616 |       "text/plain": [
 617 |        "1.0"
 618 |       ]
 619 |      },
 620 |      "execution_count": 15,
 621 |      "metadata": {},
 622 |      "output_type": "execute_result"
 623 |     }
 624 |    ],
 625 |    "source": [
 626 |     "test.index.isin(spp['sku']).mean()"
 627 |    ]
 628 |   },
 629 |   {
 630 |    "cell_type": "code",
 631 |    "execution_count": 16,
 632 |    "id": "joint-reservation",
 633 |    "metadata": {},
 634 |    "outputs": [],
 635 |    "source": [
 636 |     "prob_array = utils.pred_list_to_tweedie(spp['days_to_so'].values, phi=2., p=1.5)\n",
 637 |     "pd.set_option(\"display.max_columns\", 31)\n",
 638 |     "pd.DataFrame(prob_array).round(4).to_csv(\"19.csv.gz\", header=False, index=False, compression=\"gzip\")"
 639 |    ]
 640 |   },
 641 |   {
 642 |    "cell_type": "code",
 643 |    "execution_count": 18,
 644 |    "id": "fitting-hamilton",
 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>0</th>\n",
 669 |        "      <th>1</th>\n",
 670 |        "      <th>2</th>\n",
 671 |        "      <th>3</th>\n",
 672 |        "      <th>4</th>\n",
 673 |        "      <th>5</th>\n",
 674 |        "      <th>6</th>\n",
 675 |        "      <th>7</th>\n",
 676 |        "      <th>8</th>\n",
 677 |        "      <th>9</th>\n",
 678 |        "      <th>10</th>\n",
 679 |        "      <th>11</th>\n",
 680 |        "      <th>12</th>\n",
 681 |        "      <th>13</th>\n",
 682 |        "      <th>14</th>\n",
 683 |        "      <th>15</th>\n",
 684 |        "      <th>16</th>\n",
 685 |        "      <th>17</th>\n",
 686 |        "      <th>18</th>\n",
 687 |        "      <th>19</th>\n",
 688 |        "      <th>20</th>\n",
 689 |        "      <th>21</th>\n",
 690 |        "      <th>22</th>\n",
 691 |        "      <th>23</th>\n",
 692 |        "      <th>24</th>\n",
 693 |        "      <th>25</th>\n",
 694 |        "      <th>26</th>\n",
 695 |        "      <th>27</th>\n",
 696 |        "      <th>28</th>\n",
 697 |        "      <th>29</th>\n",
 698 |        "    </tr>\n",
 699 |        "  </thead>\n",
 700 |        "  <tbody>\n",
 701 |        "    <tr>\n",
 702 |        "      <th>0</th>\n",
 703 |        "      <td>0.0215</td>\n",
 704 |        "      <td>0.0145</td>\n",
 705 |        "      <td>0.0175</td>\n",
 706 |        "      <td>0.0204</td>\n",
 707 |        "      <td>0.0233</td>\n",
 708 |        "      <td>0.0259</td>\n",
 709 |        "      <td>0.0284</td>\n",
 710 |        "      <td>0.0307</td>\n",
 711 |        "      <td>0.0328</td>\n",
 712 |        "      <td>0.0346</td>\n",
 713 |        "      <td>0.0362</td>\n",
 714 |        "      <td>0.0375</td>\n",
 715 |        "      <td>0.0386</td>\n",
 716 |        "      <td>0.0395</td>\n",
 717 |        "      <td>0.0401</td>\n",
 718 |        "      <td>0.0405</td>\n",
 719 |        "      <td>0.0407</td>\n",
 720 |        "      <td>0.0407</td>\n",
 721 |        "      <td>0.0405</td>\n",
 722 |        "      <td>0.0402</td>\n",
 723 |        "      <td>0.0397</td>\n",
 724 |        "      <td>0.0390</td>\n",
 725 |        "      <td>0.0383</td>\n",
 726 |        "      <td>0.0374</td>\n",
 727 |        "      <td>0.0364</td>\n",
 728 |        "      <td>0.0354</td>\n",
 729 |        "      <td>0.0343</td>\n",
 730 |        "      <td>0.0331</td>\n",
 731 |        "      <td>0.0319</td>\n",
 732 |        "      <td>0.0307</td>\n",
 733 |        "    </tr>\n",
 734 |        "    <tr>\n",
 735 |        "      <th>1</th>\n",
 736 |        "      <td>0.0161</td>\n",
 737 |        "      <td>0.0111</td>\n",
 738 |        "      <td>0.0137</td>\n",
 739 |        "      <td>0.0162</td>\n",
 740 |        "      <td>0.0188</td>\n",
 741 |        "      <td>0.0213</td>\n",
 742 |        "      <td>0.0238</td>\n",
 743 |        "      <td>0.0261</td>\n",
 744 |        "      <td>0.0284</td>\n",
 745 |        "      <td>0.0305</td>\n",
 746 |        "      <td>0.0325</td>\n",
 747 |        "      <td>0.0343</td>\n",
 748 |        "      <td>0.0359</td>\n",
 749 |        "      <td>0.0373</td>\n",
 750 |        "      <td>0.0386</td>\n",
 751 |        "      <td>0.0397</td>\n",
 752 |        "      <td>0.0406</td>\n",
 753 |        "      <td>0.0413</td>\n",
 754 |        "      <td>0.0418</td>\n",
 755 |        "      <td>0.0422</td>\n",
 756 |        "      <td>0.0424</td>\n",
 757 |        "      <td>0.0424</td>\n",
 758 |        "      <td>0.0423</td>\n",
 759 |        "      <td>0.0421</td>\n",
 760 |        "      <td>0.0417</td>\n",
 761 |        "      <td>0.0412</td>\n",
 762 |        "      <td>0.0406</td>\n",
 763 |        "      <td>0.0399</td>\n",
 764 |        "      <td>0.0391</td>\n",
 765 |        "      <td>0.0383</td>\n",
 766 |        "    </tr>\n",
 767 |        "    <tr>\n",
 768 |        "      <th>2</th>\n",
 769 |        "      <td>0.0161</td>\n",
 770 |        "      <td>0.0111</td>\n",
 771 |        "      <td>0.0137</td>\n",
 772 |        "      <td>0.0162</td>\n",
 773 |        "      <td>0.0188</td>\n",
 774 |        "      <td>0.0213</td>\n",
 775 |        "      <td>0.0238</td>\n",
 776 |        "      <td>0.0261</td>\n",
 777 |        "      <td>0.0284</td>\n",
 778 |        "      <td>0.0305</td>\n",
 779 |        "      <td>0.0325</td>\n",
 780 |        "      <td>0.0343</td>\n",
 781 |        "      <td>0.0359</td>\n",
 782 |        "      <td>0.0373</td>\n",
 783 |        "      <td>0.0386</td>\n",
 784 |        "      <td>0.0397</td>\n",
 785 |        "      <td>0.0406</td>\n",
 786 |        "      <td>0.0413</td>\n",
 787 |        "      <td>0.0418</td>\n",
 788 |        "      <td>0.0422</td>\n",
 789 |        "      <td>0.0424</td>\n",
 790 |        "      <td>0.0424</td>\n",
 791 |        "      <td>0.0423</td>\n",
 792 |        "      <td>0.0421</td>\n",
 793 |        "      <td>0.0417</td>\n",
 794 |        "      <td>0.0412</td>\n",
 795 |        "      <td>0.0406</td>\n",
 796 |        "      <td>0.0399</td>\n",
 797 |        "      <td>0.0391</td>\n",
 798 |        "      <td>0.0383</td>\n",
 799 |        "    </tr>\n",
 800 |        "    <tr>\n",
 801 |        "      <th>3</th>\n",
 802 |        "      <td>0.0161</td>\n",
 803 |        "      <td>0.0111</td>\n",
 804 |        "      <td>0.0137</td>\n",
 805 |        "      <td>0.0162</td>\n",
 806 |        "      <td>0.0188</td>\n",
 807 |        "      <td>0.0213</td>\n",
 808 |        "      <td>0.0238</td>\n",
 809 |        "      <td>0.0261</td>\n",
 810 |        "      <td>0.0284</td>\n",
 811 |        "      <td>0.0305</td>\n",
 812 |        "      <td>0.0325</td>\n",
 813 |        "      <td>0.0343</td>\n",
 814 |        "      <td>0.0359</td>\n",
 815 |        "      <td>0.0373</td>\n",
 816 |        "      <td>0.0386</td>\n",
 817 |        "      <td>0.0397</td>\n",
 818 |        "      <td>0.0406</td>\n",
 819 |        "      <td>0.0413</td>\n",
 820 |        "      <td>0.0418</td>\n",
 821 |        "      <td>0.0422</td>\n",
 822 |        "      <td>0.0424</td>\n",
 823 |        "      <td>0.0424</td>\n",
 824 |        "      <td>0.0423</td>\n",
 825 |        "      <td>0.0421</td>\n",
 826 |        "      <td>0.0417</td>\n",
 827 |        "      <td>0.0412</td>\n",
 828 |        "      <td>0.0406</td>\n",
 829 |        "      <td>0.0399</td>\n",
 830 |        "      <td>0.0391</td>\n",
 831 |        "      <td>0.0383</td>\n",
 832 |        "    </tr>\n",
 833 |        "    <tr>\n",
 834 |        "      <th>4</th>\n",
 835 |        "      <td>0.0161</td>\n",
 836 |        "      <td>0.0111</td>\n",
 837 |        "      <td>0.0137</td>\n",
 838 |        "      <td>0.0162</td>\n",
 839 |        "      <td>0.0188</td>\n",
 840 |        "      <td>0.0213</td>\n",
 841 |        "      <td>0.0238</td>\n",
 842 |        "      <td>0.0261</td>\n",
 843 |        "      <td>0.0284</td>\n",
 844 |        "      <td>0.0305</td>\n",
 845 |        "      <td>0.0325</td>\n",
 846 |        "      <td>0.0343</td>\n",
 847 |        "      <td>0.0359</td>\n",
 848 |        "      <td>0.0373</td>\n",
 849 |        "      <td>0.0386</td>\n",
 850 |        "      <td>0.0397</td>\n",
 851 |        "      <td>0.0406</td>\n",
 852 |        "      <td>0.0413</td>\n",
 853 |        "      <td>0.0418</td>\n",
 854 |        "      <td>0.0422</td>\n",
 855 |        "      <td>0.0424</td>\n",
 856 |        "      <td>0.0424</td>\n",
 857 |        "      <td>0.0423</td>\n",
 858 |        "      <td>0.0421</td>\n",
 859 |        "      <td>0.0417</td>\n",
 860 |        "      <td>0.0412</td>\n",
 861 |        "      <td>0.0406</td>\n",
 862 |        "      <td>0.0399</td>\n",
 863 |        "      <td>0.0391</td>\n",
 864 |        "      <td>0.0383</td>\n",
 865 |        "    </tr>\n",
 866 |        "    <tr>\n",
 867 |        "      <th>...</th>\n",
 868 |        "      <td>...</td>\n",
 869 |        "      <td>...</td>\n",
 870 |        "      <td>...</td>\n",
 871 |        "      <td>...</td>\n",
 872 |        "      <td>...</td>\n",
 873 |        "      <td>...</td>\n",
 874 |        "      <td>...</td>\n",
 875 |        "      <td>...</td>\n",
 876 |        "      <td>...</td>\n",
 877 |        "      <td>...</td>\n",
 878 |        "      <td>...</td>\n",
 879 |        "      <td>...</td>\n",
 880 |        "      <td>...</td>\n",
 881 |        "      <td>...</td>\n",
 882 |        "      <td>...</td>\n",
 883 |        "      <td>...</td>\n",
 884 |        "      <td>...</td>\n",
 885 |        "      <td>...</td>\n",
 886 |        "      <td>...</td>\n",
 887 |        "      <td>...</td>\n",
 888 |        "      <td>...</td>\n",
 889 |        "      <td>...</td>\n",
 890 |        "      <td>...</td>\n",
 891 |        "      <td>...</td>\n",
 892 |        "      <td>...</td>\n",
 893 |        "      <td>...</td>\n",
 894 |        "      <td>...</td>\n",
 895 |        "      <td>...</td>\n",
 896 |        "      <td>...</td>\n",
 897 |        "      <td>...</td>\n",
 898 |        "    </tr>\n",
 899 |        "    <tr>\n",
 900 |        "      <th>551467</th>\n",
 901 |        "      <td>0.0161</td>\n",
 902 |        "      <td>0.0111</td>\n",
 903 |        "      <td>0.0137</td>\n",
 904 |        "      <td>0.0162</td>\n",
 905 |        "      <td>0.0188</td>\n",
 906 |        "      <td>0.0213</td>\n",
 907 |        "      <td>0.0238</td>\n",
 908 |        "      <td>0.0261</td>\n",
 909 |        "      <td>0.0284</td>\n",
 910 |        "      <td>0.0305</td>\n",
 911 |        "      <td>0.0325</td>\n",
 912 |        "      <td>0.0343</td>\n",
 913 |        "      <td>0.0359</td>\n",
 914 |        "      <td>0.0373</td>\n",
 915 |        "      <td>0.0386</td>\n",
 916 |        "      <td>0.0397</td>\n",
 917 |        "      <td>0.0406</td>\n",
 918 |        "      <td>0.0413</td>\n",
 919 |        "      <td>0.0418</td>\n",
 920 |        "      <td>0.0422</td>\n",
 921 |        "      <td>0.0424</td>\n",
 922 |        "      <td>0.0424</td>\n",
 923 |        "      <td>0.0423</td>\n",
 924 |        "      <td>0.0421</td>\n",
 925 |        "      <td>0.0417</td>\n",
 926 |        "      <td>0.0412</td>\n",
 927 |        "      <td>0.0406</td>\n",
 928 |        "      <td>0.0399</td>\n",
 929 |        "      <td>0.0391</td>\n",
 930 |        "      <td>0.0383</td>\n",
 931 |        "    </tr>\n",
 932 |        "    <tr>\n",
 933 |        "      <th>551468</th>\n",
 934 |        "      <td>0.0161</td>\n",
 935 |        "      <td>0.0111</td>\n",
 936 |        "      <td>0.0137</td>\n",
 937 |        "      <td>0.0162</td>\n",
 938 |        "      <td>0.0188</td>\n",
 939 |        "      <td>0.0213</td>\n",
 940 |        "      <td>0.0238</td>\n",
 941 |        "      <td>0.0261</td>\n",
 942 |        "      <td>0.0284</td>\n",
 943 |        "      <td>0.0305</td>\n",
 944 |        "      <td>0.0325</td>\n",
 945 |        "      <td>0.0343</td>\n",
 946 |        "      <td>0.0359</td>\n",
 947 |        "      <td>0.0373</td>\n",
 948 |        "      <td>0.0386</td>\n",
 949 |        "      <td>0.0397</td>\n",
 950 |        "      <td>0.0406</td>\n",
 951 |        "      <td>0.0413</td>\n",
 952 |        "      <td>0.0418</td>\n",
 953 |        "      <td>0.0422</td>\n",
 954 |        "      <td>0.0424</td>\n",
 955 |        "      <td>0.0424</td>\n",
 956 |        "      <td>0.0423</td>\n",
 957 |        "      <td>0.0421</td>\n",
 958 |        "      <td>0.0417</td>\n",
 959 |        "      <td>0.0412</td>\n",
 960 |        "      <td>0.0406</td>\n",
 961 |        "      <td>0.0399</td>\n",
 962 |        "      <td>0.0391</td>\n",
 963 |        "      <td>0.0383</td>\n",
 964 |        "    </tr>\n",
 965 |        "    <tr>\n",
 966 |        "      <th>551469</th>\n",
 967 |        "      <td>0.0161</td>\n",
 968 |        "      <td>0.0111</td>\n",
 969 |        "      <td>0.0137</td>\n",
 970 |        "      <td>0.0162</td>\n",
 971 |        "      <td>0.0188</td>\n",
 972 |        "      <td>0.0213</td>\n",
 973 |        "      <td>0.0238</td>\n",
 974 |        "      <td>0.0261</td>\n",
 975 |        "      <td>0.0284</td>\n",
 976 |        "      <td>0.0305</td>\n",
 977 |        "      <td>0.0325</td>\n",
 978 |        "      <td>0.0343</td>\n",
 979 |        "      <td>0.0359</td>\n",
 980 |        "      <td>0.0373</td>\n",
 981 |        "      <td>0.0386</td>\n",
 982 |        "      <td>0.0397</td>\n",
 983 |        "      <td>0.0406</td>\n",
 984 |        "      <td>0.0413</td>\n",
 985 |        "      <td>0.0418</td>\n",
 986 |        "      <td>0.0422</td>\n",
 987 |        "      <td>0.0424</td>\n",
 988 |        "      <td>0.0424</td>\n",
 989 |        "      <td>0.0423</td>\n",
 990 |        "      <td>0.0421</td>\n",
 991 |        "      <td>0.0417</td>\n",
 992 |        "      <td>0.0412</td>\n",
 993 |        "      <td>0.0406</td>\n",
 994 |        "      <td>0.0399</td>\n",
 995 |        "      <td>0.0391</td>\n",
 996 |        "      <td>0.0383</td>\n",
 997 |        "    </tr>\n",
 998 |        "    <tr>\n",
 999 |        "      <th>551470</th>\n",
1000 |        "      <td>0.0399</td>\n",
1001 |        "      <td>0.0256</td>\n",
1002 |        "      <td>0.0296</td>\n",
1003 |        "      <td>0.0331</td>\n",
1004 |        "      <td>0.0361</td>\n",
1005 |        "      <td>0.0386</td>\n",
1006 |        "      <td>0.0405</td>\n",
1007 |        "      <td>0.0419</td>\n",
1008 |        "      <td>0.0429</td>\n",
1009 |        "      <td>0.0434</td>\n",
1010 |        "      <td>0.0435</td>\n",
1011 |        "      <td>0.0432</td>\n",
1012 |        "      <td>0.0427</td>\n",
1013 |        "      <td>0.0418</td>\n",
1014 |        "      <td>0.0407</td>\n",
1015 |        "      <td>0.0394</td>\n",
1016 |        "      <td>0.0379</td>\n",
1017 |        "      <td>0.0364</td>\n",
1018 |        "      <td>0.0347</td>\n",
1019 |        "      <td>0.0330</td>\n",
1020 |        "      <td>0.0312</td>\n",
1021 |        "      <td>0.0294</td>\n",
1022 |        "      <td>0.0276</td>\n",
1023 |        "      <td>0.0259</td>\n",
1024 |        "      <td>0.0242</td>\n",
1025 |        "      <td>0.0225</td>\n",
1026 |        "      <td>0.0209</td>\n",
1027 |        "      <td>0.0193</td>\n",
1028 |        "      <td>0.0178</td>\n",
1029 |        "      <td>0.0164</td>\n",
1030 |        "    </tr>\n",
1031 |        "    <tr>\n",
1032 |        "      <th>551471</th>\n",
1033 |        "      <td>0.0161</td>\n",
1034 |        "      <td>0.0111</td>\n",
1035 |        "      <td>0.0137</td>\n",
1036 |        "      <td>0.0162</td>\n",
1037 |        "      <td>0.0188</td>\n",
1038 |        "      <td>0.0213</td>\n",
1039 |        "      <td>0.0238</td>\n",
1040 |        "      <td>0.0261</td>\n",
1041 |        "      <td>0.0284</td>\n",
1042 |        "      <td>0.0305</td>\n",
1043 |        "      <td>0.0325</td>\n",
1044 |        "      <td>0.0343</td>\n",
1045 |        "      <td>0.0359</td>\n",
1046 |        "      <td>0.0373</td>\n",
1047 |        "      <td>0.0386</td>\n",
1048 |        "      <td>0.0397</td>\n",
1049 |        "      <td>0.0406</td>\n",
1050 |        "      <td>0.0413</td>\n",
1051 |        "      <td>0.0418</td>\n",
1052 |        "      <td>0.0422</td>\n",
1053 |        "      <td>0.0424</td>\n",
1054 |        "      <td>0.0424</td>\n",
1055 |        "      <td>0.0423</td>\n",
1056 |        "      <td>0.0421</td>\n",
1057 |        "      <td>0.0417</td>\n",
1058 |        "      <td>0.0412</td>\n",
1059 |        "      <td>0.0406</td>\n",
1060 |        "      <td>0.0399</td>\n",
1061 |        "      <td>0.0391</td>\n",
1062 |        "      <td>0.0383</td>\n",
1063 |        "    </tr>\n",
1064 |        "  </tbody>\n",
1065 |        "</table>\n",
1066 |        "<p>551472 rows × 30 columns</p>\n",
1067 |        "</div>"
1068 |       ],
1069 |       "text/plain": [
1070 |        "            0       1       2       3       4       5       6       7   \\\n",
1071 |        "0       0.0215  0.0145  0.0175  0.0204  0.0233  0.0259  0.0284  0.0307   \n",
1072 |        "1       0.0161  0.0111  0.0137  0.0162  0.0188  0.0213  0.0238  0.0261   \n",
1073 |        "2       0.0161  0.0111  0.0137  0.0162  0.0188  0.0213  0.0238  0.0261   \n",
1074 |        "3       0.0161  0.0111  0.0137  0.0162  0.0188  0.0213  0.0238  0.0261   \n",
1075 |        "4       0.0161  0.0111  0.0137  0.0162  0.0188  0.0213  0.0238  0.0261   \n",
1076 |        "...        ...     ...     ...     ...     ...     ...     ...     ...   \n",
1077 |        "551467  0.0161  0.0111  0.0137  0.0162  0.0188  0.0213  0.0238  0.0261   \n",
1078 |        "551468  0.0161  0.0111  0.0137  0.0162  0.0188  0.0213  0.0238  0.0261   \n",
1079 |        "551469  0.0161  0.0111  0.0137  0.0162  0.0188  0.0213  0.0238  0.0261   \n",
1080 |        "551470  0.0399  0.0256  0.0296  0.0331  0.0361  0.0386  0.0405  0.0419   \n",
1081 |        "551471  0.0161  0.0111  0.0137  0.0162  0.0188  0.0213  0.0238  0.0261   \n",
1082 |        "\n",
1083 |        "            8       9       10      11      12      13      14      15  \\\n",
1084 |        "0       0.0328  0.0346  0.0362  0.0375  0.0386  0.0395  0.0401  0.0405   \n",
1085 |        "1       0.0284  0.0305  0.0325  0.0343  0.0359  0.0373  0.0386  0.0397   \n",
1086 |        "2       0.0284  0.0305  0.0325  0.0343  0.0359  0.0373  0.0386  0.0397   \n",
1087 |        "3       0.0284  0.0305  0.0325  0.0343  0.0359  0.0373  0.0386  0.0397   \n",
1088 |        "4       0.0284  0.0305  0.0325  0.0343  0.0359  0.0373  0.0386  0.0397   \n",
1089 |        "...        ...     ...     ...     ...     ...     ...     ...     ...   \n",
1090 |        "551467  0.0284  0.0305  0.0325  0.0343  0.0359  0.0373  0.0386  0.0397   \n",
1091 |        "551468  0.0284  0.0305  0.0325  0.0343  0.0359  0.0373  0.0386  0.0397   \n",
1092 |        "551469  0.0284  0.0305  0.0325  0.0343  0.0359  0.0373  0.0386  0.0397   \n",
1093 |        "551470  0.0429  0.0434  0.0435  0.0432  0.0427  0.0418  0.0407  0.0394   \n",
1094 |        "551471  0.0284  0.0305  0.0325  0.0343  0.0359  0.0373  0.0386  0.0397   \n",
1095 |        "\n",
1096 |        "            16      17      18      19      20      21      22      23  \\\n",
1097 |        "0       0.0407  0.0407  0.0405  0.0402  0.0397  0.0390  0.0383  0.0374   \n",
1098 |        "1       0.0406  0.0413  0.0418  0.0422  0.0424  0.0424  0.0423  0.0421   \n",
1099 |        "2       0.0406  0.0413  0.0418  0.0422  0.0424  0.0424  0.0423  0.0421   \n",
1100 |        "3       0.0406  0.0413  0.0418  0.0422  0.0424  0.0424  0.0423  0.0421   \n",
1101 |        "4       0.0406  0.0413  0.0418  0.0422  0.0424  0.0424  0.0423  0.0421   \n",
1102 |        "...        ...     ...     ...     ...     ...     ...     ...     ...   \n",
1103 |        "551467  0.0406  0.0413  0.0418  0.0422  0.0424  0.0424  0.0423  0.0421   \n",
1104 |        "551468  0.0406  0.0413  0.0418  0.0422  0.0424  0.0424  0.0423  0.0421   \n",
1105 |        "551469  0.0406  0.0413  0.0418  0.0422  0.0424  0.0424  0.0423  0.0421   \n",
1106 |        "551470  0.0379  0.0364  0.0347  0.0330  0.0312  0.0294  0.0276  0.0259   \n",
1107 |        "551471  0.0406  0.0413  0.0418  0.0422  0.0424  0.0424  0.0423  0.0421   \n",
1108 |        "\n",
1109 |        "            24      25      26      27      28      29  \n",
1110 |        "0       0.0364  0.0354  0.0343  0.0331  0.0319  0.0307  \n",
1111 |        "1       0.0417  0.0412  0.0406  0.0399  0.0391  0.0383  \n",
1112 |        "2       0.0417  0.0412  0.0406  0.0399  0.0391  0.0383  \n",
1113 |        "3       0.0417  0.0412  0.0406  0.0399  0.0391  0.0383  \n",
1114 |        "4       0.0417  0.0412  0.0406  0.0399  0.0391  0.0383  \n",
1115 |        "...        ...     ...     ...     ...     ...     ...  \n",
1116 |        "551467  0.0417  0.0412  0.0406  0.0399  0.0391  0.0383  \n",
1117 |        "551468  0.0417  0.0412  0.0406  0.0399  0.0391  0.0383  \n",
1118 |        "551469  0.0417  0.0412  0.0406  0.0399  0.0391  0.0383  \n",
1119 |        "551470  0.0242  0.0225  0.0209  0.0193  0.0178  0.0164  \n",
1120 |        "551471  0.0417  0.0412  0.0406  0.0399  0.0391  0.0383  \n",
1121 |        "\n",
1122 |        "[551472 rows x 30 columns]"
1123 |       ]
1124 |      },
1125 |      "execution_count": 18,
1126 |      "metadata": {},
1127 |      "output_type": "execute_result"
1128 |     }
1129 |    ],
1130 |    "source": [
1131 |     "pd.read_csv(\"19.csv.gz\",header=None)#.sum(axis=1)"
1132 |    ]
1133 |   },
1134 |   {
1135 |    "cell_type": "code",
1136 |    "execution_count": null,
1137 |    "id": "eleven-intelligence",
1138 |    "metadata": {},
1139 |    "outputs": [],
1140 |    "source": []
1141 |   }
1142 |  ],
1143 |  "metadata": {
1144 |   "kernelspec": {
1145 |    "display_name": "Python 3",
1146 |    "language": "python",
1147 |    "name": "python3"
1148 |   },
1149 |   "language_info": {
1150 |    "codemirror_mode": {
1151 |     "name": "ipython",
1152 |     "version": 3
1153 |    },
1154 |    "file_extension": ".py",
1155 |    "mimetype": "text/x-python",
1156 |    "name": "python",
1157 |    "nbconvert_exporter": "python",
1158 |    "pygments_lexer": "ipython3",
1159 |    "version": "3.8.5"
1160 |   }
1161 |  },
1162 |  "nbformat": 4,
1163 |  "nbformat_minor": 5
1164 | }
1165 | 


--------------------------------------------------------------------------------
/meli2021/32_baseline_yt.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 4,
  6 |    "id": "75638528-4609-4ee2-93c9-28538579e471",
  7 |    "metadata": {},
  8 |    "outputs": [],
  9 |    "source": [
 10 |     "import pandas as pd\n",
 11 |     "import numpy as np\n",
 12 |     "import utils\n",
 13 |     "\n",
 14 |     "from sklearn.model_selection import GroupKFold, KFold\n",
 15 |     "from sklearn.linear_model import LinearRegression\n",
 16 |     "from sklearn.ensemble import RandomForestRegressor\n",
 17 |     "from sklearn.metrics import mean_squared_error\n",
 18 |     "from xgboost import XGBRegressor\n",
 19 |     "import tweedie\n",
 20 |     "\n",
 21 |     "from importlib import reload\n",
 22 |     "reload(utils)\n",
 23 |     "from skopt import gp_minimize\n",
 24 |     "%matplotlib inline"
 25 |    ]
 26 |   },
 27 |   {
 28 |    "cell_type": "code",
 29 |    "execution_count": 5,
 30 |    "id": "3be069d6-c17e-43a2-8252-2acf5e9c26ad",
 31 |    "metadata": {},
 32 |    "outputs": [],
 33 |    "source": [
 34 |     "train = pd.read_parquet(\"./train/0.parquet\")\n",
 35 |     "train['date'] = pd.to_datetime(train['date'])\n",
 36 |     "train['fold'] = train['date'].dt.month"
 37 |    ]
 38 |   },
 39 |   {
 40 |    "cell_type": "code",
 41 |    "execution_count": 6,
 42 |    "id": "smaller-boulder",
 43 |    "metadata": {},
 44 |    "outputs": [],
 45 |    "source": [
 46 |     "test = pd.read_csv(\"test_data.csv\", index_col=0).squeeze()"
 47 |    ]
 48 |   },
 49 |   {
 50 |    "cell_type": "code",
 51 |    "execution_count": 7,
 52 |    "id": "finnish-canadian",
 53 |    "metadata": {},
 54 |    "outputs": [
 55 |     {
 56 |      "data": {
 57 |       "text/plain": [
 58 |        "sku                                int64\n",
 59 |        "date                      datetime64[ns]\n",
 60 |        "sold_quantity                      int64\n",
 61 |        "current_price                    float64\n",
 62 |        "currency                          object\n",
 63 |        "listing_type                      object\n",
 64 |        "shipping_logistic_type            object\n",
 65 |        "shipping_payment                  object\n",
 66 |        "minutes_active                   float64\n",
 67 |        "item_domain_id                    object\n",
 68 |        "site_id                           object\n",
 69 |        "fold                               int64\n",
 70 |        "dtype: object"
 71 |       ]
 72 |      },
 73 |      "execution_count": 7,
 74 |      "metadata": {},
 75 |      "output_type": "execute_result"
 76 |     }
 77 |    ],
 78 |    "source": [
 79 |     "train.dtypes"
 80 |    ]
 81 |   },
 82 |   {
 83 |    "cell_type": "code",
 84 |    "execution_count": 8,
 85 |    "id": "fc10499b-4c46-4f7e-a88e-a4233fb05504",
 86 |    "metadata": {},
 87 |    "outputs": [
 88 |     {
 89 |      "data": {
 90 |       "text/html": [
 91 |        "<div>\n",
 92 |        "<style scoped>\n",
 93 |        "    .dataframe tbody tr th:only-of-type {\n",
 94 |        "        vertical-align: middle;\n",
 95 |        "    }\n",
 96 |        "\n",
 97 |        "    .dataframe tbody tr th {\n",
 98 |        "        vertical-align: top;\n",
 99 |        "    }\n",
100 |        "\n",
101 |        "    .dataframe thead th {\n",
102 |        "        text-align: right;\n",
103 |        "    }\n",
104 |        "</style>\n",
105 |        "<table border=\"1\" class=\"dataframe\">\n",
106 |        "  <thead>\n",
107 |        "    <tr style=\"text-align: right;\">\n",
108 |        "      <th></th>\n",
109 |        "      <th>sku</th>\n",
110 |        "      <th>date</th>\n",
111 |        "      <th>sold_quantity</th>\n",
112 |        "      <th>current_price</th>\n",
113 |        "      <th>currency</th>\n",
114 |        "      <th>listing_type</th>\n",
115 |        "      <th>shipping_logistic_type</th>\n",
116 |        "      <th>shipping_payment</th>\n",
117 |        "      <th>minutes_active</th>\n",
118 |        "      <th>item_domain_id</th>\n",
119 |        "      <th>site_id</th>\n",
120 |        "      <th>fold</th>\n",
121 |        "    </tr>\n",
122 |        "  </thead>\n",
123 |        "  <tbody>\n",
124 |        "    <tr>\n",
125 |        "      <th>0</th>\n",
126 |        "      <td>464801</td>\n",
127 |        "      <td>2021-02-01</td>\n",
128 |        "      <td>0</td>\n",
129 |        "      <td>156.78</td>\n",
130 |        "      <td>REA</td>\n",
131 |        "      <td>classic</td>\n",
132 |        "      <td>fulfillment</td>\n",
133 |        "      <td>free_shipping</td>\n",
134 |        "      <td>1440.0</td>\n",
135 |        "      <td>MLB-NEBULIZERS</td>\n",
136 |        "      <td>MLB</td>\n",
137 |        "      <td>2</td>\n",
138 |        "    </tr>\n",
139 |        "    <tr>\n",
140 |        "      <th>1</th>\n",
141 |        "      <td>464801</td>\n",
142 |        "      <td>2021-02-02</td>\n",
143 |        "      <td>0</td>\n",
144 |        "      <td>156.78</td>\n",
145 |        "      <td>REA</td>\n",
146 |        "      <td>classic</td>\n",
147 |        "      <td>fulfillment</td>\n",
148 |        "      <td>free_shipping</td>\n",
149 |        "      <td>1440.0</td>\n",
150 |        "      <td>MLB-NEBULIZERS</td>\n",
151 |        "      <td>MLB</td>\n",
152 |        "      <td>2</td>\n",
153 |        "    </tr>\n",
154 |        "    <tr>\n",
155 |        "      <th>2</th>\n",
156 |        "      <td>464801</td>\n",
157 |        "      <td>2021-02-03</td>\n",
158 |        "      <td>0</td>\n",
159 |        "      <td>156.78</td>\n",
160 |        "      <td>REA</td>\n",
161 |        "      <td>classic</td>\n",
162 |        "      <td>fulfillment</td>\n",
163 |        "      <td>free_shipping</td>\n",
164 |        "      <td>1440.0</td>\n",
165 |        "      <td>MLB-NEBULIZERS</td>\n",
166 |        "      <td>MLB</td>\n",
167 |        "      <td>2</td>\n",
168 |        "    </tr>\n",
169 |        "    <tr>\n",
170 |        "      <th>3</th>\n",
171 |        "      <td>464801</td>\n",
172 |        "      <td>2021-02-04</td>\n",
173 |        "      <td>0</td>\n",
174 |        "      <td>156.78</td>\n",
175 |        "      <td>REA</td>\n",
176 |        "      <td>classic</td>\n",
177 |        "      <td>fulfillment</td>\n",
178 |        "      <td>free_shipping</td>\n",
179 |        "      <td>1440.0</td>\n",
180 |        "      <td>MLB-NEBULIZERS</td>\n",
181 |        "      <td>MLB</td>\n",
182 |        "      <td>2</td>\n",
183 |        "    </tr>\n",
184 |        "    <tr>\n",
185 |        "      <th>4</th>\n",
186 |        "      <td>464801</td>\n",
187 |        "      <td>2021-02-05</td>\n",
188 |        "      <td>1</td>\n",
189 |        "      <td>156.78</td>\n",
190 |        "      <td>REA</td>\n",
191 |        "      <td>classic</td>\n",
192 |        "      <td>fulfillment</td>\n",
193 |        "      <td>free_shipping</td>\n",
194 |        "      <td>1440.0</td>\n",
195 |        "      <td>MLB-NEBULIZERS</td>\n",
196 |        "      <td>MLB</td>\n",
197 |        "      <td>2</td>\n",
198 |        "    </tr>\n",
199 |        "  </tbody>\n",
200 |        "</table>\n",
201 |        "</div>"
202 |       ],
203 |       "text/plain": [
204 |        "      sku       date  sold_quantity  current_price currency listing_type  \\\n",
205 |        "0  464801 2021-02-01              0         156.78      REA      classic   \n",
206 |        "1  464801 2021-02-02              0         156.78      REA      classic   \n",
207 |        "2  464801 2021-02-03              0         156.78      REA      classic   \n",
208 |        "3  464801 2021-02-04              0         156.78      REA      classic   \n",
209 |        "4  464801 2021-02-05              1         156.78      REA      classic   \n",
210 |        "\n",
211 |        "  shipping_logistic_type shipping_payment  minutes_active  item_domain_id  \\\n",
212 |        "0            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
213 |        "1            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
214 |        "2            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
215 |        "3            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
216 |        "4            fulfillment    free_shipping          1440.0  MLB-NEBULIZERS   \n",
217 |        "\n",
218 |        "  site_id  fold  \n",
219 |        "0     MLB     2  \n",
220 |        "1     MLB     2  \n",
221 |        "2     MLB     2  \n",
222 |        "3     MLB     2  \n",
223 |        "4     MLB     2  "
224 |       ]
225 |      },
226 |      "execution_count": 8,
227 |      "metadata": {},
228 |      "output_type": "execute_result"
229 |     }
230 |    ],
231 |    "source": [
232 |     "train.head()"
233 |    ]
234 |   },
235 |   {
236 |    "cell_type": "code",
237 |    "execution_count": 9,
238 |    "id": "monthly-general",
239 |    "metadata": {},
240 |    "outputs": [],
241 |    "source": [
242 |     "def gen_tr_ts():\n",
243 |     "    for fold in [2,3]:\n",
244 |     "        ts = train[train['fold'] != fold]['date'].max()\n",
245 |     "        ts = train[(train['fold'] != fold) & (train['date'] == ts)].index\n",
246 |     "        yield train.index[train['fold'] == fold], ts, fold\n",
247 |     "\n",
248 |     "    "
249 |    ]
250 |   },
251 |   {
252 |    "cell_type": "code",
253 |    "execution_count": 10,
254 |    "id": "bdb25520-ed36-4c4c-a9fe-23387cb3f918",
255 |    "metadata": {},
256 |    "outputs": [],
257 |    "source": [
258 |     "\n",
259 |     "def gen_p(na=30., power=1.5):\n",
260 |     "    dts = list()\n",
261 |     "    for tr,ts, fold in gen_tr_ts():\n",
262 |     "\n",
263 |     "        ## EVAL\n",
264 |     "        pp = train[train['fold'] != fold][['sku', 'date', 'sold_quantity']]\n",
265 |     "        pp['stock'] = pp['sku'].map(test)\n",
266 |     "        pp = pp.sort_values([\"sku\",\"date\"])\n",
267 |     "        pp['cumulative_y'] = pp.groupby(\"sku\")['sold_quantity'].cumsum()\n",
268 |     "\n",
269 |     "        pp = pp.dropna(subset=['stock'])\n",
270 |     "        pp['stockout_y'] = pp['cumulative_y'] >= pp['stock']\n",
271 |     "\n",
272 |     "        first_so_y = pp[pp['stockout_y']].groupby(\"sku\").first()\n",
273 |     "        days_to_so_y = (first_so_y[\"date\"] - pp[\"date\"].min()) / np.timedelta64(1, 'D')\n",
274 |     "        days_to_so_y = days_to_so_y.reindex(pp['sku'].unique()).fillna(na).astype(int).clip(1)\n",
275 |     "        dts.append(days_to_so_y)\n",
276 |     "\n",
277 |     "    m = utils.pred_list_to_distro(dts[0], wei=False, total_days=max(na, 30), phi=2, power=power)\n",
278 |     "    f = utils.pred_list_to_distro(dts[1].reindex(dts[0].index).fillna(dts[0]), wei=False, total_days=max(na, 30), phi=2, power=power)\n",
279 |     "\n",
280 |     "    m = pd.DataFrame(m,index=dts[0].index)\n",
281 |     "    f = pd.DataFrame(f,index=dts[0].index)\n",
282 |     "\n",
283 |     "    p = (m + f)/2\n",
284 |     "\n",
285 |     "    p = p.div(p.sum(axis=1), axis=0)\n",
286 |     "\n",
287 |     "    p = p.loc[test.index]\n",
288 |     "    \n",
289 |     "    p = p.round(4)\n",
290 |     "\n",
291 |     "    return p"
292 |    ]
293 |   },
294 |   {
295 |    "cell_type": "code",
296 |    "execution_count": 11,
297 |    "id": "outer-certificate",
298 |    "metadata": {},
299 |    "outputs": [],
300 |    "source": [
301 |     "r = {i:gen_p(30, power=i/10) for i in range(11,20,1)}\n"
302 |    ]
303 |   },
304 |   {
305 |    "cell_type": "code",
306 |    "execution_count": 52,
307 |    "id": "1834ec6c-f4e4-4272-8b08-f6a3cbadb705",
308 |    "metadata": {},
309 |    "outputs": [
310 |     {
311 |      "data": {
312 |       "text/plain": [
313 |        "<matplotlib.axes._subplots.AxesSubplot at 0x7f7ef8c79310>"
314 |       ]
315 |      },
316 |      "execution_count": 52,
317 |      "metadata": {},
318 |      "output_type": "execute_result"
319 |     },
320 |     {
321 |      "data": {
322 |       "image/png": "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\n",
323 |       "text/plain": [
324 |        "<Figure size 432x288 with 1 Axes>"
325 |       ]
326 |      },
327 |      "metadata": {
328 |       "needs_background": "light"
329 |      },
330 |      "output_type": "display_data"
331 |     }
332 |    ],
333 |    "source": [
334 |     "r[11].mean(axis=0).plot()"
335 |    ]
336 |   },
337 |   {
338 |    "cell_type": "code",
339 |    "execution_count": 53,
340 |    "id": "bd7c0392-db7e-4361-a7ed-64cf9a949504",
341 |    "metadata": {},
342 |    "outputs": [
343 |     {
344 |      "data": {
345 |       "text/plain": [
346 |        "<matplotlib.axes._subplots.AxesSubplot at 0x7f7f18da1a00>"
347 |       ]
348 |      },
349 |      "execution_count": 53,
350 |      "metadata": {},
351 |      "output_type": "execute_result"
352 |     },
353 |     {
354 |      "data": {
355 |       "image/png": "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\n",
356 |       "text/plain": [
357 |        "<Figure size 432x288 with 1 Axes>"
358 |       ]
359 |      },
360 |      "metadata": {
361 |       "needs_background": "light"
362 |      },
363 |      "output_type": "display_data"
364 |     }
365 |    ],
366 |    "source": [
367 |     "r[12].mean(axis=0).plot()"
368 |    ]
369 |   },
370 |   {
371 |    "cell_type": "code",
372 |    "execution_count": 13,
373 |    "id": "7b22e456-318b-4ab0-bf80-31807d2f268c",
374 |    "metadata": {},
375 |    "outputs": [
376 |     {
377 |      "data": {
378 |       "text/plain": [
379 |        "<matplotlib.axes._subplots.AxesSubplot at 0x7fb628658670>"
380 |       ]
381 |      },
382 |      "execution_count": 13,
383 |      "metadata": {},
384 |      "output_type": "execute_result"
385 |     },
386 |     {
387 |      "data": {
388 |       "image/png": "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\n",
389 |       "text/plain": [
390 |        "<Figure size 432x288 with 1 Axes>"
391 |       ]
392 |      },
393 |      "metadata": {
394 |       "needs_background": "light"
395 |      },
396 |      "output_type": "display_data"
397 |     }
398 |    ],
399 |    "source": [
400 |     "r[13].mean(axis=0).plot()"
401 |    ]
402 |   },
403 |   {
404 |    "cell_type": "code",
405 |    "execution_count": 59,
406 |    "id": "2efa8e12-e62a-463a-82e4-bdca3babdcd3",
407 |    "metadata": {},
408 |    "outputs": [],
409 |    "source": [
410 |     "# phi=2, power=1.3, na=30\n",
411 |     "r[13].round(4).to_csv(\"32b.csv.gz\", header=False, index=False, compression=\"gzip\")\n",
412 |     "# LB 4.94"
413 |    ]
414 |   },
415 |   {
416 |    "cell_type": "code",
417 |    "execution_count": null,
418 |    "id": "57e747a1-1d2b-4339-8be8-ffdd1c6fb0a3",
419 |    "metadata": {},
420 |    "outputs": [],
421 |    "source": [
422 |     "# BONUS\n",
423 |     "# xgb 19 + baseline 32 / 2 - LB 4.44"
424 |    ]
425 |   }
426 |  ],
427 |  "metadata": {
428 |   "kernelspec": {
429 |    "display_name": "Python 3",
430 |    "language": "python",
431 |    "name": "python3"
432 |   },
433 |   "language_info": {
434 |    "codemirror_mode": {
435 |     "name": "ipython",
436 |     "version": 3
437 |    },
438 |    "file_extension": ".py",
439 |    "mimetype": "text/x-python",
440 |    "name": "python",
441 |    "nbconvert_exporter": "python",
442 |    "pygments_lexer": "ipython3",
443 |    "version": "3.8.5"
444 |   }
445 |  },
446 |  "nbformat": 4,
447 |  "nbformat_minor": 5
448 | }
449 | 


--------------------------------------------------------------------------------
/meli2021/61_active_model_yt.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 1,
  6 |    "id": "7fe10a1f-3c36-4466-89c5-5dc2622daf87",
  7 |    "metadata": {},
  8 |    "outputs": [],
  9 |    "source": [
 10 |     "import pandas as pd\n",
 11 |     "%matplotlib inline\n",
 12 |     "import numpy as np\n",
 13 |     "from matplotlib import pyplot as plt"
 14 |    ]
 15 |   },
 16 |   {
 17 |    "cell_type": "code",
 18 |    "execution_count": null,
 19 |    "id": "c171946b-edfc-4a51-90fe-cc483428228f",
 20 |    "metadata": {},
 21 |    "outputs": [],
 22 |    "source": [
 23 |     "Y -> how many days until this product becomes active?"
 24 |    ]
 25 |   },
 26 |   {
 27 |    "cell_type": "code",
 28 |    "execution_count": 2,
 29 |    "id": "84e02a3e-1668-40a5-83cd-3029def45fd8",
 30 |    "metadata": {},
 31 |    "outputs": [],
 32 |    "source": [
 33 |     "test = pd.read_csv(\"test_data.csv\").set_index(\"sku\").squeeze()\n",
 34 |     "train = pd.read_parquet(\"./train/0.parquet\")\n",
 35 |     "train['date'] = pd.to_datetime(train['date'])\n",
 36 |     "cats = ['item_domain_id', 'currency', 'listing_type', 'shipping_logistic_type', 'shipping_payment', 'site_id']\n",
 37 |     "for cat in cats:\n",
 38 |     "    train[cat] = train[cat].astype(\"category\").cat.codes"
 39 |    ]
 40 |   },
 41 |   {
 42 |    "cell_type": "code",
 43 |    "execution_count": 3,
 44 |    "id": "8ca0f337-197d-4a22-9191-7bef0b33b705",
 45 |    "metadata": {},
 46 |    "outputs": [],
 47 |    "source": [
 48 |     "train.loc[train[\"minutes_active\"] == 0, \"active\"] = 0\n",
 49 |     "train[\"active\"] = train[\"active\"].fillna(1)"
 50 |    ]
 51 |   },
 52 |   {
 53 |    "cell_type": "code",
 54 |    "execution_count": 4,
 55 |    "id": "311a92d9-55b1-484d-a53b-ef755681d75c",
 56 |    "metadata": {},
 57 |    "outputs": [
 58 |     {
 59 |      "data": {
 60 |       "text/html": [
 61 |        "<div>\n",
 62 |        "<style scoped>\n",
 63 |        "    .dataframe tbody tr th:only-of-type {\n",
 64 |        "        vertical-align: middle;\n",
 65 |        "    }\n",
 66 |        "\n",
 67 |        "    .dataframe tbody tr th {\n",
 68 |        "        vertical-align: top;\n",
 69 |        "    }\n",
 70 |        "\n",
 71 |        "    .dataframe thead th {\n",
 72 |        "        text-align: right;\n",
 73 |        "    }\n",
 74 |        "</style>\n",
 75 |        "<table border=\"1\" class=\"dataframe\">\n",
 76 |        "  <thead>\n",
 77 |        "    <tr style=\"text-align: right;\">\n",
 78 |        "      <th></th>\n",
 79 |        "      <th>sku</th>\n",
 80 |        "      <th>shipping_logistic_type</th>\n",
 81 |        "      <th>shipping_payment</th>\n",
 82 |        "      <th>listing_type</th>\n",
 83 |        "      <th>currency</th>\n",
 84 |        "      <th>current_price</th>\n",
 85 |        "      <th>item_domain_id</th>\n",
 86 |        "      <th>site_id</th>\n",
 87 |        "      <th>days_to_active</th>\n",
 88 |        "      <th>days_since_inactive</th>\n",
 89 |        "    </tr>\n",
 90 |        "  </thead>\n",
 91 |        "  <tbody>\n",
 92 |        "    <tr>\n",
 93 |        "      <th>sku</th>\n",
 94 |        "      <td>1.000000</td>\n",
 95 |        "      <td>-0.000341</td>\n",
 96 |        "      <td>-0.002559</td>\n",
 97 |        "      <td>-0.000866</td>\n",
 98 |        "      <td>-0.002091</td>\n",
 99 |        "      <td>0.002278</td>\n",
100 |        "      <td>-0.001439</td>\n",
101 |        "      <td>0.000258</td>\n",
102 |        "      <td>-0.001643</td>\n",
103 |        "      <td>-0.001465</td>\n",
104 |        "    </tr>\n",
105 |        "    <tr>\n",
106 |        "      <th>shipping_logistic_type</th>\n",
107 |        "      <td>-0.000341</td>\n",
108 |        "      <td>1.000000</td>\n",
109 |        "      <td>0.074183</td>\n",
110 |        "      <td>0.040746</td>\n",
111 |        "      <td>0.050840</td>\n",
112 |        "      <td>-0.089014</td>\n",
113 |        "      <td>-0.031270</td>\n",
114 |        "      <td>-0.019540</td>\n",
115 |        "      <td>-0.009456</td>\n",
116 |        "      <td>0.075642</td>\n",
117 |        "    </tr>\n",
118 |        "    <tr>\n",
119 |        "      <th>shipping_payment</th>\n",
120 |        "      <td>-0.002559</td>\n",
121 |        "      <td>0.074183</td>\n",
122 |        "      <td>1.000000</td>\n",
123 |        "      <td>-0.074338</td>\n",
124 |        "      <td>0.107770</td>\n",
125 |        "      <td>-0.620856</td>\n",
126 |        "      <td>-0.118603</td>\n",
127 |        "      <td>-0.111873</td>\n",
128 |        "      <td>-0.012127</td>\n",
129 |        "      <td>-0.009789</td>\n",
130 |        "    </tr>\n",
131 |        "    <tr>\n",
132 |        "      <th>listing_type</th>\n",
133 |        "      <td>-0.000866</td>\n",
134 |        "      <td>0.040746</td>\n",
135 |        "      <td>-0.074338</td>\n",
136 |        "      <td>1.000000</td>\n",
137 |        "      <td>0.152052</td>\n",
138 |        "      <td>-0.035744</td>\n",
139 |        "      <td>0.072715</td>\n",
140 |        "      <td>0.096491</td>\n",
141 |        "      <td>0.029645</td>\n",
142 |        "      <td>-0.007888</td>\n",
143 |        "    </tr>\n",
144 |        "    <tr>\n",
145 |        "      <th>currency</th>\n",
146 |        "      <td>-0.002091</td>\n",
147 |        "      <td>0.050840</td>\n",
148 |        "      <td>0.107770</td>\n",
149 |        "      <td>0.152052</td>\n",
150 |        "      <td>1.000000</td>\n",
151 |        "      <td>-0.657856</td>\n",
152 |        "      <td>-0.422581</td>\n",
153 |        "      <td>-0.472765</td>\n",
154 |        "      <td>-0.014007</td>\n",
155 |        "      <td>-0.011959</td>\n",
156 |        "    </tr>\n",
157 |        "    <tr>\n",
158 |        "      <th>current_price</th>\n",
159 |        "      <td>0.002278</td>\n",
160 |        "      <td>-0.089014</td>\n",
161 |        "      <td>-0.620856</td>\n",
162 |        "      <td>-0.035744</td>\n",
163 |        "      <td>-0.657856</td>\n",
164 |        "      <td>1.000000</td>\n",
165 |        "      <td>0.267825</td>\n",
166 |        "      <td>0.274570</td>\n",
167 |        "      <td>0.014325</td>\n",
168 |        "      <td>0.006590</td>\n",
169 |        "    </tr>\n",
170 |        "    <tr>\n",
171 |        "      <th>item_domain_id</th>\n",
172 |        "      <td>-0.001439</td>\n",
173 |        "      <td>-0.031270</td>\n",
174 |        "      <td>-0.118603</td>\n",
175 |        "      <td>0.072715</td>\n",
176 |        "      <td>-0.422581</td>\n",
177 |        "      <td>0.267825</td>\n",
178 |        "      <td>1.000000</td>\n",
179 |        "      <td>0.893819</td>\n",
180 |        "      <td>0.040363</td>\n",
181 |        "      <td>0.022034</td>\n",
182 |        "    </tr>\n",
183 |        "    <tr>\n",
184 |        "      <th>site_id</th>\n",
185 |        "      <td>0.000258</td>\n",
186 |        "      <td>-0.019540</td>\n",
187 |        "      <td>-0.111873</td>\n",
188 |        "      <td>0.096491</td>\n",
189 |        "      <td>-0.472765</td>\n",
190 |        "      <td>0.274570</td>\n",
191 |        "      <td>0.893819</td>\n",
192 |        "      <td>1.000000</td>\n",
193 |        "      <td>0.038815</td>\n",
194 |        "      <td>0.023721</td>\n",
195 |        "    </tr>\n",
196 |        "    <tr>\n",
197 |        "      <th>days_to_active</th>\n",
198 |        "      <td>-0.001643</td>\n",
199 |        "      <td>-0.009456</td>\n",
200 |        "      <td>-0.012127</td>\n",
201 |        "      <td>0.029645</td>\n",
202 |        "      <td>-0.014007</td>\n",
203 |        "      <td>0.014325</td>\n",
204 |        "      <td>0.040363</td>\n",
205 |        "      <td>0.038815</td>\n",
206 |        "      <td>1.000000</td>\n",
207 |        "      <td>-0.118599</td>\n",
208 |        "    </tr>\n",
209 |        "    <tr>\n",
210 |        "      <th>days_since_inactive</th>\n",
211 |        "      <td>-0.001465</td>\n",
212 |        "      <td>0.075642</td>\n",
213 |        "      <td>-0.009789</td>\n",
214 |        "      <td>-0.007888</td>\n",
215 |        "      <td>-0.011959</td>\n",
216 |        "      <td>0.006590</td>\n",
217 |        "      <td>0.022034</td>\n",
218 |        "      <td>0.023721</td>\n",
219 |        "      <td>-0.118599</td>\n",
220 |        "      <td>1.000000</td>\n",
221 |        "    </tr>\n",
222 |        "  </tbody>\n",
223 |        "</table>\n",
224 |        "</div>"
225 |       ],
226 |       "text/plain": [
227 |        "                             sku  shipping_logistic_type  shipping_payment  \\\n",
228 |        "sku                     1.000000               -0.000341         -0.002559   \n",
229 |        "shipping_logistic_type -0.000341                1.000000          0.074183   \n",
230 |        "shipping_payment       -0.002559                0.074183          1.000000   \n",
231 |        "listing_type           -0.000866                0.040746         -0.074338   \n",
232 |        "currency               -0.002091                0.050840          0.107770   \n",
233 |        "current_price           0.002278               -0.089014         -0.620856   \n",
234 |        "item_domain_id         -0.001439               -0.031270         -0.118603   \n",
235 |        "site_id                 0.000258               -0.019540         -0.111873   \n",
236 |        "days_to_active         -0.001643               -0.009456         -0.012127   \n",
237 |        "days_since_inactive    -0.001465                0.075642         -0.009789   \n",
238 |        "\n",
239 |        "                        listing_type  currency  current_price  item_domain_id  \\\n",
240 |        "sku                        -0.000866 -0.002091       0.002278       -0.001439   \n",
241 |        "shipping_logistic_type      0.040746  0.050840      -0.089014       -0.031270   \n",
242 |        "shipping_payment           -0.074338  0.107770      -0.620856       -0.118603   \n",
243 |        "listing_type                1.000000  0.152052      -0.035744        0.072715   \n",
244 |        "currency                    0.152052  1.000000      -0.657856       -0.422581   \n",
245 |        "current_price              -0.035744 -0.657856       1.000000        0.267825   \n",
246 |        "item_domain_id              0.072715 -0.422581       0.267825        1.000000   \n",
247 |        "site_id                     0.096491 -0.472765       0.274570        0.893819   \n",
248 |        "days_to_active              0.029645 -0.014007       0.014325        0.040363   \n",
249 |        "days_since_inactive        -0.007888 -0.011959       0.006590        0.022034   \n",
250 |        "\n",
251 |        "                         site_id  days_to_active  days_since_inactive  \n",
252 |        "sku                     0.000258       -0.001643            -0.001465  \n",
253 |        "shipping_logistic_type -0.019540       -0.009456             0.075642  \n",
254 |        "shipping_payment       -0.111873       -0.012127            -0.009789  \n",
255 |        "listing_type            0.096491        0.029645            -0.007888  \n",
256 |        "currency               -0.472765       -0.014007            -0.011959  \n",
257 |        "current_price           0.274570        0.014325             0.006590  \n",
258 |        "item_domain_id          0.893819        0.040363             0.022034  \n",
259 |        "site_id                 1.000000        0.038815             0.023721  \n",
260 |        "days_to_active          0.038815        1.000000            -0.118599  \n",
261 |        "days_since_inactive     0.023721       -0.118599             1.000000  "
262 |       ]
263 |      },
264 |      "execution_count": 4,
265 |      "metadata": {},
266 |      "output_type": "execute_result"
267 |     }
268 |    ],
269 |    "source": [
270 |     "act = train[train['active'] == 1][['sku', 'date']].sort_values(\"date\")\n",
271 |     "act['active_date'] = act['date']\n",
272 |     "inact = train[train['active'] == 0][['sku', 'date', 'shipping_logistic_type', 'shipping_payment', \n",
273 |     "                                                               'listing_type', 'currency', 'current_price', 'item_domain_id', 'site_id']].sort_values(\"date\")\n",
274 |     "all_ = pd.merge_asof(inact, act, on=['date'], direction='forward', by=['sku']).dropna(subset=['active_date'])\n",
275 |     "all_['days_to_active'] = (all_['active_date'] - all_['date']) / np.timedelta64(1,'D')\n",
276 |     "all_['days_since_inactive'] = (all_['date'] - all_.groupby(\"sku\")[\"date\"].transform(\"min\")) / np.timedelta64(1,'D')\n",
277 |     "y = all_['days_to_active'].copy()\n",
278 |     "all_.corr(method='spearman')"
279 |    ]
280 |   },
281 |   {
282 |    "cell_type": "code",
283 |    "execution_count": 5,
284 |    "id": "bf5ab194-5311-4008-807e-8024a0e7b599",
285 |    "metadata": {},
286 |    "outputs": [],
287 |    "source": [
288 |     "import xgboost as xgb\n",
289 |     "from sklearn.metrics import mean_squared_error"
290 |    ]
291 |   },
292 |   {
293 |    "cell_type": "code",
294 |    "execution_count": 6,
295 |    "id": "141e9fc5-9b36-4933-bf31-317f5fa88f47",
296 |    "metadata": {},
297 |    "outputs": [
298 |     {
299 |      "data": {
300 |       "text/plain": [
301 |        "7.689729948430816"
302 |       ]
303 |      },
304 |      "execution_count": 6,
305 |      "metadata": {},
306 |      "output_type": "execute_result"
307 |     }
308 |    ],
309 |    "source": [
310 |     "Xtr = all_.loc[all_['date'] < \"2021-03-01\", ['days_since_inactive', 'current_price'] + cats]\n",
311 |     "Xval = all_.loc[all_['date'] >= \"2021-03-01\", ['days_since_inactive', 'current_price'] + cats]\n",
312 |     "\n",
313 |     "ytr = y[all_['date'] < \"2021-03-01\"]\n",
314 |     "yval = y[all_['date'] >= \"2021-03-01\"]\n",
315 |     "\n",
316 |     "#mdl = DecisionTreeRegressor(max_depth=3)\n",
317 |     "#mdl = RandomForestRegressor(n_estimators=100, max_depth=1, random_state=0, n_jobs=6)\n",
318 |     "mdl = xgb.XGBRegressor(n_estimators=100, max_depth=3, learning_rate=0.1, random_state=0, n_jobs=6, tree_method='hist')\n",
319 |     "mdl.fit(Xtr, ytr)\n",
320 |     "p = mdl.predict(Xval)\n",
321 |     "np.sqrt(mean_squared_error(yval, p))"
322 |    ]
323 |   },
324 |   {
325 |    "cell_type": "code",
326 |    "execution_count": 7,
327 |    "id": "f7df5b24-3d9d-4412-864b-b0f32da1cbd9",
328 |    "metadata": {},
329 |    "outputs": [
330 |     {
331 |      "data": {
332 |       "text/plain": [
333 |        "<matplotlib.axes._subplots.AxesSubplot at 0x7f7f5004a670>"
334 |       ]
335 |      },
336 |      "execution_count": 7,
337 |      "metadata": {},
338 |      "output_type": "execute_result"
339 |     },
340 |     {
341 |      "data": {
342 |       "image/png": "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\n",
343 |       "text/plain": [
344 |        "<Figure size 1080x720 with 1 Axes>"
345 |       ]
346 |      },
347 |      "metadata": {
348 |       "needs_background": "light"
349 |      },
350 |      "output_type": "display_data"
351 |     }
352 |    ],
353 |    "source": [
354 |     "Xval['p'] = p\n",
355 |     "Xval['error'] = Xval['p'] - yval\n",
356 |     "\n",
357 |     "Xval.groupby(\"days_since_inactive\")['error'].mean().plot.bar(figsize=(15,10))"
358 |    ]
359 |   },
360 |   {
361 |    "cell_type": "code",
362 |    "execution_count": 8,
363 |    "id": "638ead4d-b6bb-4c3f-a11e-ac1ee378a80c",
364 |    "metadata": {},
365 |    "outputs": [],
366 |    "source": [
367 |     "all_t = train[train['active'] == 0].copy()\n",
368 |     "\n",
369 |     "all_t['days_since_inactive'] = (all_t['date'] - all_t.groupby(\"sku\")[\"date\"].transform(\"min\")) / np.timedelta64(1,'D')\n",
370 |     "all_t = all_t.groupby(\"sku\").last()\n",
371 |     "all_t = all_t[all_t['date'] == \"2021-03-31\"].copy()"
372 |    ]
373 |   },
374 |   {
375 |    "cell_type": "code",
376 |    "execution_count": 9,
377 |    "id": "f17ca4b5-3352-4747-88ce-b6de1fc0b480",
378 |    "metadata": {},
379 |    "outputs": [
380 |     {
381 |      "data": {
382 |       "text/html": [
383 |        "<div>\n",
384 |        "<style scoped>\n",
385 |        "    .dataframe tbody tr th:only-of-type {\n",
386 |        "        vertical-align: middle;\n",
387 |        "    }\n",
388 |        "\n",
389 |        "    .dataframe tbody tr th {\n",
390 |        "        vertical-align: top;\n",
391 |        "    }\n",
392 |        "\n",
393 |        "    .dataframe thead th {\n",
394 |        "        text-align: right;\n",
395 |        "    }\n",
396 |        "</style>\n",
397 |        "<table border=\"1\" class=\"dataframe\">\n",
398 |        "  <thead>\n",
399 |        "    <tr style=\"text-align: right;\">\n",
400 |        "      <th></th>\n",
401 |        "      <th>date</th>\n",
402 |        "      <th>sold_quantity</th>\n",
403 |        "      <th>current_price</th>\n",
404 |        "      <th>currency</th>\n",
405 |        "      <th>listing_type</th>\n",
406 |        "      <th>shipping_logistic_type</th>\n",
407 |        "      <th>shipping_payment</th>\n",
408 |        "      <th>minutes_active</th>\n",
409 |        "      <th>item_domain_id</th>\n",
410 |        "      <th>site_id</th>\n",
411 |        "      <th>active</th>\n",
412 |        "      <th>days_since_inactive</th>\n",
413 |        "    </tr>\n",
414 |        "    <tr>\n",
415 |        "      <th>sku</th>\n",
416 |        "      <th></th>\n",
417 |        "      <th></th>\n",
418 |        "      <th></th>\n",
419 |        "      <th></th>\n",
420 |        "      <th></th>\n",
421 |        "      <th></th>\n",
422 |        "      <th></th>\n",
423 |        "      <th></th>\n",
424 |        "      <th></th>\n",
425 |        "      <th></th>\n",
426 |        "      <th></th>\n",
427 |        "      <th></th>\n",
428 |        "    </tr>\n",
429 |        "  </thead>\n",
430 |        "  <tbody>\n",
431 |        "    <tr>\n",
432 |        "      <th>4</th>\n",
433 |        "      <td>2021-03-31</td>\n",
434 |        "      <td>0</td>\n",
435 |        "      <td>118.00</td>\n",
436 |        "      <td>3</td>\n",
437 |        "      <td>1</td>\n",
438 |        "      <td>2</td>\n",
439 |        "      <td>0</td>\n",
440 |        "      <td>0.0</td>\n",
441 |        "      <td>5263</td>\n",
442 |        "      <td>1</td>\n",
443 |        "      <td>0.0</td>\n",
444 |        "      <td>54.0</td>\n",
445 |        "    </tr>\n",
446 |        "    <tr>\n",
447 |        "      <th>8</th>\n",
448 |        "      <td>2021-03-31</td>\n",
449 |        "      <td>0</td>\n",
450 |        "      <td>58.49</td>\n",
451 |        "      <td>2</td>\n",
452 |        "      <td>1</td>\n",
453 |        "      <td>2</td>\n",
454 |        "      <td>1</td>\n",
455 |        "      <td>0.0</td>\n",
456 |        "      <td>6116</td>\n",
457 |        "      <td>2</td>\n",
458 |        "      <td>0.0</td>\n",
459 |        "      <td>41.0</td>\n",
460 |        "    </tr>\n",
461 |        "    <tr>\n",
462 |        "      <th>9</th>\n",
463 |        "      <td>2021-03-31</td>\n",
464 |        "      <td>0</td>\n",
465 |        "      <td>199.00</td>\n",
466 |        "      <td>2</td>\n",
467 |        "      <td>1</td>\n",
468 |        "      <td>2</td>\n",
469 |        "      <td>0</td>\n",
470 |        "      <td>0.0</td>\n",
471 |        "      <td>7991</td>\n",
472 |        "      <td>2</td>\n",
473 |        "      <td>0.0</td>\n",
474 |        "      <td>6.0</td>\n",
475 |        "    </tr>\n",
476 |        "    <tr>\n",
477 |        "      <th>11</th>\n",
478 |        "      <td>2021-03-31</td>\n",
479 |        "      <td>0</td>\n",
480 |        "      <td>109.90</td>\n",
481 |        "      <td>3</td>\n",
482 |        "      <td>0</td>\n",
483 |        "      <td>2</td>\n",
484 |        "      <td>0</td>\n",
485 |        "      <td>0.0</td>\n",
486 |        "      <td>3645</td>\n",
487 |        "      <td>1</td>\n",
488 |        "      <td>0.0</td>\n",
489 |        "      <td>19.0</td>\n",
490 |        "    </tr>\n",
491 |        "    <tr>\n",
492 |        "      <th>13</th>\n",
493 |        "      <td>2021-03-31</td>\n",
494 |        "      <td>0</td>\n",
495 |        "      <td>474.05</td>\n",
496 |        "      <td>2</td>\n",
497 |        "      <td>1</td>\n",
498 |        "      <td>2</td>\n",
499 |        "      <td>0</td>\n",
500 |        "      <td>0.0</td>\n",
501 |        "      <td>5658</td>\n",
502 |        "      <td>2</td>\n",
503 |        "      <td>0.0</td>\n",
504 |        "      <td>0.0</td>\n",
505 |        "    </tr>\n",
506 |        "    <tr>\n",
507 |        "      <th>...</th>\n",
508 |        "      <td>...</td>\n",
509 |        "      <td>...</td>\n",
510 |        "      <td>...</td>\n",
511 |        "      <td>...</td>\n",
512 |        "      <td>...</td>\n",
513 |        "      <td>...</td>\n",
514 |        "      <td>...</td>\n",
515 |        "      <td>...</td>\n",
516 |        "      <td>...</td>\n",
517 |        "      <td>...</td>\n",
518 |        "      <td>...</td>\n",
519 |        "      <td>...</td>\n",
520 |        "    </tr>\n",
521 |        "    <tr>\n",
522 |        "      <th>660897</th>\n",
523 |        "      <td>2021-03-31</td>\n",
524 |        "      <td>0</td>\n",
525 |        "      <td>79.00</td>\n",
526 |        "      <td>2</td>\n",
527 |        "      <td>0</td>\n",
528 |        "      <td>2</td>\n",
529 |        "      <td>1</td>\n",
530 |        "      <td>0.0</td>\n",
531 |        "      <td>6201</td>\n",
532 |        "      <td>2</td>\n",
533 |        "      <td>0.0</td>\n",
534 |        "      <td>58.0</td>\n",
535 |        "    </tr>\n",
536 |        "    <tr>\n",
537 |        "      <th>660904</th>\n",
538 |        "      <td>2021-03-31</td>\n",
539 |        "      <td>0</td>\n",
540 |        "      <td>98.90</td>\n",
541 |        "      <td>3</td>\n",
542 |        "      <td>1</td>\n",
543 |        "      <td>2</td>\n",
544 |        "      <td>1</td>\n",
545 |        "      <td>0.0</td>\n",
546 |        "      <td>3678</td>\n",
547 |        "      <td>1</td>\n",
548 |        "      <td>0.0</td>\n",
549 |        "      <td>24.0</td>\n",
550 |        "    </tr>\n",
551 |        "    <tr>\n",
552 |        "      <th>660907</th>\n",
553 |        "      <td>2021-03-31</td>\n",
554 |        "      <td>0</td>\n",
555 |        "      <td>24.69</td>\n",
556 |        "      <td>3</td>\n",
557 |        "      <td>1</td>\n",
558 |        "      <td>2</td>\n",
559 |        "      <td>1</td>\n",
560 |        "      <td>0.0</td>\n",
561 |        "      <td>4703</td>\n",
562 |        "      <td>1</td>\n",
563 |        "      <td>0.0</td>\n",
564 |        "      <td>56.0</td>\n",
565 |        "    </tr>\n",
566 |        "    <tr>\n",
567 |        "      <th>660910</th>\n",
568 |        "      <td>2021-03-31</td>\n",
569 |        "      <td>0</td>\n",
570 |        "      <td>480.58</td>\n",
571 |        "      <td>2</td>\n",
572 |        "      <td>1</td>\n",
573 |        "      <td>2</td>\n",
574 |        "      <td>0</td>\n",
575 |        "      <td>0.0</td>\n",
576 |        "      <td>7878</td>\n",
577 |        "      <td>2</td>\n",
578 |        "      <td>0.0</td>\n",
579 |        "      <td>42.0</td>\n",
580 |        "    </tr>\n",
581 |        "    <tr>\n",
582 |        "      <th>660915</th>\n",
583 |        "      <td>2021-03-31</td>\n",
584 |        "      <td>0</td>\n",
585 |        "      <td>99.99</td>\n",
586 |        "      <td>2</td>\n",
587 |        "      <td>0</td>\n",
588 |        "      <td>2</td>\n",
589 |        "      <td>1</td>\n",
590 |        "      <td>0.0</td>\n",
591 |        "      <td>7994</td>\n",
592 |        "      <td>2</td>\n",
593 |        "      <td>0.0</td>\n",
594 |        "      <td>58.0</td>\n",
595 |        "    </tr>\n",
596 |        "  </tbody>\n",
597 |        "</table>\n",
598 |        "<p>181606 rows × 12 columns</p>\n",
599 |        "</div>"
600 |       ],
601 |       "text/plain": [
602 |        "             date  sold_quantity  current_price  currency  listing_type  \\\n",
603 |        "sku                                                                       \n",
604 |        "4      2021-03-31              0         118.00         3             1   \n",
605 |        "8      2021-03-31              0          58.49         2             1   \n",
606 |        "9      2021-03-31              0         199.00         2             1   \n",
607 |        "11     2021-03-31              0         109.90         3             0   \n",
608 |        "13     2021-03-31              0         474.05         2             1   \n",
609 |        "...           ...            ...            ...       ...           ...   \n",
610 |        "660897 2021-03-31              0          79.00         2             0   \n",
611 |        "660904 2021-03-31              0          98.90         3             1   \n",
612 |        "660907 2021-03-31              0          24.69         3             1   \n",
613 |        "660910 2021-03-31              0         480.58         2             1   \n",
614 |        "660915 2021-03-31              0          99.99         2             0   \n",
615 |        "\n",
616 |        "        shipping_logistic_type  shipping_payment  minutes_active  \\\n",
617 |        "sku                                                                \n",
618 |        "4                            2                 0             0.0   \n",
619 |        "8                            2                 1             0.0   \n",
620 |        "9                            2                 0             0.0   \n",
621 |        "11                           2                 0             0.0   \n",
622 |        "13                           2                 0             0.0   \n",
623 |        "...                        ...               ...             ...   \n",
624 |        "660897                       2                 1             0.0   \n",
625 |        "660904                       2                 1             0.0   \n",
626 |        "660907                       2                 1             0.0   \n",
627 |        "660910                       2                 0             0.0   \n",
628 |        "660915                       2                 1             0.0   \n",
629 |        "\n",
630 |        "        item_domain_id  site_id  active  days_since_inactive  \n",
631 |        "sku                                                           \n",
632 |        "4                 5263        1     0.0                 54.0  \n",
633 |        "8                 6116        2     0.0                 41.0  \n",
634 |        "9                 7991        2     0.0                  6.0  \n",
635 |        "11                3645        1     0.0                 19.0  \n",
636 |        "13                5658        2     0.0                  0.0  \n",
637 |        "...                ...      ...     ...                  ...  \n",
638 |        "660897            6201        2     0.0                 58.0  \n",
639 |        "660904            3678        1     0.0                 24.0  \n",
640 |        "660907            4703        1     0.0                 56.0  \n",
641 |        "660910            7878        2     0.0                 42.0  \n",
642 |        "660915            7994        2     0.0                 58.0  \n",
643 |        "\n",
644 |        "[181606 rows x 12 columns]"
645 |       ]
646 |      },
647 |      "execution_count": 9,
648 |      "metadata": {},
649 |      "output_type": "execute_result"
650 |     }
651 |    ],
652 |    "source": [
653 |     "all_t"
654 |    ]
655 |   },
656 |   {
657 |    "cell_type": "code",
658 |    "execution_count": 10,
659 |    "id": "a8513a7c-3e3d-46ec-8725-b9076633ee7a",
660 |    "metadata": {},
661 |    "outputs": [],
662 |    "source": [
663 |     "X = all_.loc[:, ['days_since_inactive', 'current_price'] + cats]\n",
664 |     "Xt = all_t.loc[:, ['days_since_inactive', 'current_price'] + cats]\n",
665 |     "\n",
666 |     "mdl = xgb.XGBRegressor(n_estimators=100, max_depth=3, learning_rate=0.1, random_state=0, n_jobs=6, tree_method='hist')\n",
667 |     "mdl.fit(X, y)\n",
668 |     "p = mdl.predict(Xt)"
669 |    ]
670 |   },
671 |   {
672 |    "cell_type": "code",
673 |    "execution_count": 11,
674 |    "id": "96be5acc-dc34-4d44-9322-fc452364666c",
675 |    "metadata": {},
676 |    "outputs": [],
677 |    "source": [
678 |     "p2 = pd.Series(p.round(), index=Xt.index).reindex(test.index).dropna().astype(int)"
679 |    ]
680 |   },
681 |   {
682 |    "cell_type": "code",
683 |    "execution_count": 12,
684 |    "id": "42bf0345-5db8-4da9-8657-b570024386d7",
685 |    "metadata": {},
686 |    "outputs": [
687 |     {
688 |      "data": {
689 |       "text/html": [
690 |        "<div>\n",
691 |        "<style scoped>\n",
692 |        "    .dataframe tbody tr th:only-of-type {\n",
693 |        "        vertical-align: middle;\n",
694 |        "    }\n",
695 |        "\n",
696 |        "    .dataframe tbody tr th {\n",
697 |        "        vertical-align: top;\n",
698 |        "    }\n",
699 |        "\n",
700 |        "    .dataframe thead th {\n",
701 |        "        text-align: right;\n",
702 |        "    }\n",
703 |        "</style>\n",
704 |        "<table border=\"1\" class=\"dataframe\">\n",
705 |        "  <thead>\n",
706 |        "    <tr style=\"text-align: right;\">\n",
707 |        "      <th></th>\n",
708 |        "      <th>days_since_inactive</th>\n",
709 |        "      <th>current_price</th>\n",
710 |        "      <th>item_domain_id</th>\n",
711 |        "      <th>currency</th>\n",
712 |        "      <th>listing_type</th>\n",
713 |        "      <th>shipping_logistic_type</th>\n",
714 |        "      <th>shipping_payment</th>\n",
715 |        "      <th>site_id</th>\n",
716 |        "    </tr>\n",
717 |        "    <tr>\n",
718 |        "      <th>sku</th>\n",
719 |        "      <th></th>\n",
720 |        "      <th></th>\n",
721 |        "      <th></th>\n",
722 |        "      <th></th>\n",
723 |        "      <th></th>\n",
724 |        "      <th></th>\n",
725 |        "      <th></th>\n",
726 |        "      <th></th>\n",
727 |        "    </tr>\n",
728 |        "  </thead>\n",
729 |        "  <tbody>\n",
730 |        "    <tr>\n",
731 |        "      <th>4</th>\n",
732 |        "      <td>54.0</td>\n",
733 |        "      <td>118.00</td>\n",
734 |        "      <td>5263</td>\n",
735 |        "      <td>3</td>\n",
736 |        "      <td>1</td>\n",
737 |        "      <td>2</td>\n",
738 |        "      <td>0</td>\n",
739 |        "      <td>1</td>\n",
740 |        "    </tr>\n",
741 |        "    <tr>\n",
742 |        "      <th>8</th>\n",
743 |        "      <td>41.0</td>\n",
744 |        "      <td>58.49</td>\n",
745 |        "      <td>6116</td>\n",
746 |        "      <td>2</td>\n",
747 |        "      <td>1</td>\n",
748 |        "      <td>2</td>\n",
749 |        "      <td>1</td>\n",
750 |        "      <td>2</td>\n",
751 |        "    </tr>\n",
752 |        "    <tr>\n",
753 |        "      <th>9</th>\n",
754 |        "      <td>6.0</td>\n",
755 |        "      <td>199.00</td>\n",
756 |        "      <td>7991</td>\n",
757 |        "      <td>2</td>\n",
758 |        "      <td>1</td>\n",
759 |        "      <td>2</td>\n",
760 |        "      <td>0</td>\n",
761 |        "      <td>2</td>\n",
762 |        "    </tr>\n",
763 |        "    <tr>\n",
764 |        "      <th>11</th>\n",
765 |        "      <td>19.0</td>\n",
766 |        "      <td>109.90</td>\n",
767 |        "      <td>3645</td>\n",
768 |        "      <td>3</td>\n",
769 |        "      <td>0</td>\n",
770 |        "      <td>2</td>\n",
771 |        "      <td>0</td>\n",
772 |        "      <td>1</td>\n",
773 |        "    </tr>\n",
774 |        "    <tr>\n",
775 |        "      <th>13</th>\n",
776 |        "      <td>0.0</td>\n",
777 |        "      <td>474.05</td>\n",
778 |        "      <td>5658</td>\n",
779 |        "      <td>2</td>\n",
780 |        "      <td>1</td>\n",
781 |        "      <td>2</td>\n",
782 |        "      <td>0</td>\n",
783 |        "      <td>2</td>\n",
784 |        "    </tr>\n",
785 |        "  </tbody>\n",
786 |        "</table>\n",
787 |        "</div>"
788 |       ],
789 |       "text/plain": [
790 |        "     days_since_inactive  current_price  item_domain_id  currency  \\\n",
791 |        "sku                                                                 \n",
792 |        "4                   54.0         118.00            5263         3   \n",
793 |        "8                   41.0          58.49            6116         2   \n",
794 |        "9                    6.0         199.00            7991         2   \n",
795 |        "11                  19.0         109.90            3645         3   \n",
796 |        "13                   0.0         474.05            5658         2   \n",
797 |        "\n",
798 |        "     listing_type  shipping_logistic_type  shipping_payment  site_id  \n",
799 |        "sku                                                                   \n",
800 |        "4               1                       2                 0        1  \n",
801 |        "8               1                       2                 1        2  \n",
802 |        "9               1                       2                 0        2  \n",
803 |        "11              0                       2                 0        1  \n",
804 |        "13              1                       2                 0        2  "
805 |       ]
806 |      },
807 |      "execution_count": 12,
808 |      "metadata": {},
809 |      "output_type": "execute_result"
810 |     }
811 |    ],
812 |    "source": [
813 |     "Xt.head()"
814 |    ]
815 |   },
816 |   {
817 |    "cell_type": "code",
818 |    "execution_count": 14,
819 |    "id": "b6880938-d9bf-456c-b8dd-c5f60dd293cd",
820 |    "metadata": {},
821 |    "outputs": [
822 |     {
823 |      "data": {
824 |       "text/plain": [
825 |        "sku\n",
826 |        "431262     2\n",
827 |        "94157      9\n",
828 |        "394886     9\n",
829 |        "434156     9\n",
830 |        "197550    12\n",
831 |        "          ..\n",
832 |        "575227    10\n",
833 |        "470249    10\n",
834 |        "24226     10\n",
835 |        "297331    10\n",
836 |        "511077    10\n",
837 |        "Length: 97692, dtype: int64"
838 |       ]
839 |      },
840 |      "execution_count": 14,
841 |      "metadata": {},
842 |      "output_type": "execute_result"
843 |     }
844 |    ],
845 |    "source": [
846 |     "p2"
847 |    ]
848 |   },
849 |   {
850 |    "cell_type": "code",
851 |    "execution_count": 13,
852 |    "id": "14642acd-37f0-420c-9951-b90ce6b6f195",
853 |    "metadata": {},
854 |    "outputs": [
855 |     {
856 |      "name": "stdout",
857 |      "output_type": "stream",
858 |      "text": [
859 |       "(551472, 30)\n",
860 |       "CPU times: user 2min 10s, sys: 836 ms, total: 2min 10s\n",
861 |       "Wall time: 2min 10s\n"
862 |      ]
863 |     }
864 |    ],
865 |    "source": [
866 |     "%%time\n",
867 |     "sub = pd.read_csv(\"45d.csv.gz\", header=None) # 4.31\n",
868 |     "sub_ = sub.copy()\n",
869 |     "sub_.index = test.index\n",
870 |     "\n",
871 |     "for sku in p2.index:\n",
872 |     "    s = sub_.loc[sku].copy()\n",
873 |     "    days = p2.loc[sku]\n",
874 |     "    #print(s)\n",
875 |     "    s.iloc[:days] = s.iloc[:days]*0.5\n",
876 |     "    s = s / s.sum()\n",
877 |     "    sub_.loc[sku, :] = s\n",
878 |     "print(sub_.shape)\n",
879 |     "sub_.round(4).to_csv(\"61byt.csv.gz\", header=False, index=False, compression=\"gzip\")\n",
880 |     "# LB 4.2772"
881 |    ]
882 |   },
883 |   {
884 |    "cell_type": "code",
885 |    "execution_count": null,
886 |    "id": "c8e7e49f-4e84-4a50-a324-76ca6abb9641",
887 |    "metadata": {},
888 |    "outputs": [],
889 |    "source": []
890 |   }
891 |  ],
892 |  "metadata": {
893 |   "kernelspec": {
894 |    "display_name": "Python 3",
895 |    "language": "python",
896 |    "name": "python3"
897 |   },
898 |   "language_info": {
899 |    "codemirror_mode": {
900 |     "name": "ipython",
901 |     "version": 3
902 |    },
903 |    "file_extension": ".py",
904 |    "mimetype": "text/x-python",
905 |    "name": "python",
906 |    "nbconvert_exporter": "python",
907 |    "pygments_lexer": "ipython3",
908 |    "version": "3.8.5"
909 |   }
910 |  },
911 |  "nbformat": 4,
912 |  "nbformat_minor": 5
913 | }
914 | 


--------------------------------------------------------------------------------
/meli2021/README:
--------------------------------------------------------------------------------
 1 | Solutions to #MeliDataChallenge 2021
 2 | 
 3 | August 2nd, 2021 -> 19_xgb_barely_tuned gives you LB 4.64 (5th place)
 4 | August 3rd, 2021 -> 32_baseline gives you LB 4.94 (7th place)
 5 | 
 6 | If you average the predictions from 19 and 32 you get LB 4.44 (3rd place)
 7 | 
 8 | August 9th, 2021 -> 61_active applied to the average above should give ~ LB 4.30 (4th place)
 9 | 
10 | Video Tutorials: https://www.youtube.com/playlist?list=PLV_itENB3unp-g1tgybj5-gs_4FGL8aA8
11 | 
12 | 


--------------------------------------------------------------------------------
/meli2021/utils.py:
--------------------------------------------------------------------------------
  1 | import numpy as np
  2 | import pandas as pd
  3 | import tweedie
  4 | import scipy.stats as st
  5 | 
  6 | def pred_list_to_prob_array(pred_list, cumulative=False, total_days=30):
  7 |     prob_array = np.zeros((pred_list.shape[0], total_days))
  8 |     pred_list = np.clip(pred_list, 1, total_days)
  9 |     for row, e in enumerate(pred_list):
 10 |         if cumulative:
 11 |             prob_array[row, int(e-1)] = 1.
 12 |         else:
 13 |             prob_array[row, int(e-1):] = 1.
 14 |         
 15 |     if cumulative:
 16 |         prob_array = prob_array+1e-4 
 17 |         prob_array = np.divide(prob_array, prob_array.sum(axis=1).reshape(-1,1))
 18 |         prob_array = prob_array.cumsum(axis=1)
 19 | 
 20 |     return prob_array
 21 | 
 22 | def pred_list_to_prob_array_mc(pred_list, total_days=30):
 23 |     prob_array = np.zeros((pred_list.shape[0], total_days))
 24 |     pred_list = np.clip(pred_list, 1, total_days)
 25 |     for row, e in enumerate(pred_list):
 26 |         prob_array[row, int(e):] = 1.
 27 |         
 28 |     return prob_array
 29 | 
 30 | def rps(y, p, probs=False, total_days=30):
 31 |     y_array = pred_list_to_prob_array(y, total_days=total_days)
 32 |     if probs:
 33 |         p_array = p.cumsum(axis=1)
 34 |     else:
 35 |         p_array = pred_list_to_prob_array(p, cumulative=True, total_days=total_days)
 36 |     return ((p_array - y_array)**2).sum(axis=1).mean()
 37 | 
 38 | 
 39 | def rps_mc(y, p, probs=False, total_days=30):
 40 |     y_array = pred_list_to_prob_array_mc(y, total_days=total_days)
 41 |     if probs:
 42 |         p_array = p.cumsum(axis=1)
 43 |     return ((p_array - y_array)**2).sum(axis=1).mean()
 44 | 
 45 | def rps_raw(y, p, probs=False):
 46 |     y_array = pred_list_to_prob_array(y)
 47 |     if probs:
 48 |         p_array = p.cumsum(axis=1)
 49 |     else:
 50 |         p_array = pred_list_to_prob_array(p, cumulative=True)
 51 |     return ((p_array - y_array)**2).sum(axis=1)
 52 | 
 53 | 
 54 | def pred_list_to_tweedie(pred_list, phi=1, p=1.5):
 55 |     # has a bug in the first day, it's the wrong probability, but it's worse without the bug
 56 |     distros = dict()
 57 |     for mu in range(1,31):
 58 |         distros[mu] = [tweedie.tweedie(p=p, mu=mu, phi=phi).cdf(days) for days in range(1,31,1)]
 59 |         distros[mu][1:] = np.diff(distros[mu])
 60 |         distros[mu] = np.round(distros[mu] / np.sum(distros[mu]), 4)
 61 |     
 62 |     prob_array = np.zeros((pred_list.shape[0], 30))
 63 | 
 64 |     for row, mu in enumerate(pred_list):
 65 |         prob_array[row, :] = distros[mu]#.cumsum()
 66 |         #prob_array[row, -1] = 1.
 67 | 
 68 |     return prob_array
 69 | 
 70 | 
 71 | 
 72 | def pred_list_to_distro(pred_list, wei=False, total_days=30, phi=2, power=1.5):
 73 |     distros = dict()
 74 |     for mu in range(1,total_days+1):
 75 |         if wei:
 76 |             distros[mu] = [st.norm.cdf(days, loc=mu, scale=1) for days in range(0,total_days+1,1)]
 77 |         else:
 78 |             distros[mu] = [tweedie.tweedie(p=power, mu=mu, phi=phi).cdf(days) for days in range(0,total_days+1,1)]
 79 |         #distros[mu] = [st.lognorm.cdf(days, s=0.5, loc=mu, scale=0.5) for days in range(0,31,1)]
 80 |         #distros[mu] = [st.expon.cdf(days, loc=mu, scale=0.01) for days in range(0,31,1)]
 81 |         #distros[mu] = [st.gengamma.cdf(days, loc=mu, scale=1, a=mu, c=1) for days in range(1,31,1)]
 82 |         if np.sum(distros[mu]) > 0:
 83 |             distros[mu] = np.diff(distros[mu])
 84 |             distros[mu] = np.round(distros[mu] / np.sum(distros[mu]), 4)
 85 |         else:
 86 |             distros[mu] = distros[mu][1:]
 87 |             distros[mu][-1] = 1
 88 |         
 89 |     
 90 |     prob_array = np.zeros((pred_list.shape[0], total_days))
 91 | 
 92 |     for row, mu in enumerate(pred_list):
 93 |         prob_array[row, :] = distros[mu]#.cumsum()
 94 |         #prob_array[row, -1] = 1.
 95 | 
 96 |     return prob_array
 97 | 
 98 | def pred_list_to_distro_smooth(pred_list, total_days=30, phi=2, power=1.5, smooth_factor=0.3):
 99 |     distros = dict()
100 |     for mu in range(1,total_days+1):
101 |         distros[mu] = [tweedie.tweedie(p=power, mu=mu, phi=phi).cdf(days) for days in range(0,total_days+1,1)]
102 |         if np.sum(distros[mu]) > 0:
103 |             distros[mu] = np.diff(distros[mu])
104 |             distros[mu] = np.round(distros[mu] / np.sum(distros[mu]), 4)
105 |         else:
106 |             distros[mu] = distros[mu][1:]
107 |             distros[mu][-1] = 1
108 |         
109 |     
110 |     prob_array = np.zeros((pred_list.shape[0], total_days))
111 | 
112 |     for row, mu in enumerate(pred_list):
113 |         if mu == 1:
114 |             prob_array[row, :] = (1-smooth_factor)*distros[mu] + smooth_factor*distros[mu+1]
115 |         elif mu == total_days:
116 |             prob_array[row, :] = smooth_factor*distros[mu-1] + (1-smooth_factor)*distros[mu]
117 |         else:
118 |             prob_array[row, :] = (smooth_factor/2)*distros[mu-1] + (1-smooth_factor)*distros[mu] + (smooth_factor/2)*distros[mu+1]
119 | 
120 |     return prob_array


--------------------------------------------------------------------------------
/multiple_time_series/README:
--------------------------------------------------------------------------------
1 | Notebooks for https://youtu.be/RRd2wzMRpOc
2 | 


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