├── .Module-01b-Explore-Regression.ipynb.icloud
├── .gitignore
├── Exercise.ipynb
├── LICENSE
├── Module-00-Installation-check.ipynb
├── Module-01a-Frame-Regression.ipynb
├── Module-01c-Model-Regression.ipynb
├── Module-02a-Intuition-Logistic.ipynb
├── Module-02b-Explore-Logistic.ipynb
├── Module-02c-Model-Logistic.ipynb
├── Module-03a-Intuition-Trees.ipynb
├── Module-03b-Model-Trees.ipynb
├── Module-03c-Model-Evaluation.ipynb
├── Module-03d-Model-Bagging.ipynb
├── Module-03e-Model-RandomForest.ipynb
├── Module-03f-Model-Boosting.ipynb
├── Module-03g-Model-HyperParameterOpt.ipynb
├── Module-04a-Regression-Basic.ipynb
├── Module-05a-ML-Pipeline.ipynb
├── README.md
├── curriculum.md
├── data
├── cars_small.csv
├── creditRisk-tree.xlsx
├── creditRisk.csv
├── historical_loan.csv
├── housing_test.csv
├── housing_train.csv
├── loan_data.csv
└── loan_data_clean.csv
├── environment.yml
├── img
├── bias_variance.png
├── confusion_matrix.jpg
├── confusion_matrix2.png
├── cross_validation.png
├── generalisation_error.png
├── linear_models.png
├── logistic-curve.png
├── logistic_regression.png
├── model_complexity.png
├── model_complexity_error.png
├── model_selection.png
├── overfitting.png
├── precision_recall.png
├── random_forest.png
├── regression_error.png
├── regularization.png
├── roc-curves.png
├── simple_complex.png
├── tree_titanic.png
└── validation.png
├── installation.md
├── modelvis-local.py
├── outline.md
├── outline.pdf
├── pre-requisites.md
├── reference
├── .Module-01b-reference.ipynb.icloud
├── Module-01a-reference.ipynb
├── Module-01c-reference.ipynb
├── Module-02b-reference.ipynb
└── Module-02c-reference.ipynb
├── schedule.md
├── tree.dot
├── tree2.dot
└── tree_3.dot
/.Module-01b-Explore-Regression.ipynb.icloud:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/amitkaps/applied-machine-learning/9752d1cd19a530b36c81632c291a6607ab3049d3/.Module-01b-Explore-Regression.ipynb.icloud
--------------------------------------------------------------------------------
/.gitignore:
--------------------------------------------------------------------------------
1 | .ipynb_checkpoints/
2 | __pycache__/
3 | Credit Risk Modeling - creating train and test.ipynb
4 | data/videoGames.csv
5 |
6 |
--------------------------------------------------------------------------------
/LICENSE:
--------------------------------------------------------------------------------
1 | MIT License
2 |
3 | Copyright (c) 2017 Amit Kapoor
4 |
5 | Permission is hereby granted, free of charge, to any person obtaining a copy
6 | of this software and associated documentation files (the "Software"), to deal
7 | in the Software without restriction, including without limitation the rights
8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9 | copies of the Software, and to permit persons to whom the Software is
10 | furnished to do so, subject to the following conditions:
11 |
12 | The above copyright notice and this permission notice shall be included in all
13 | copies or substantial portions of the Software.
14 |
15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 | SOFTWARE.
22 |
--------------------------------------------------------------------------------
/Module-00-Installation-check.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "markdown",
5 | "metadata": {},
6 | "source": [
7 | "# Installation Check\n",
8 | "\n",
9 | "This notebook is just to check we have all the libraries installationed and ready to go"
10 | ]
11 | },
12 | {
13 | "cell_type": "code",
14 | "execution_count": 1,
15 | "metadata": {},
16 | "outputs": [],
17 | "source": [
18 | "import importlib"
19 | ]
20 | },
21 | {
22 | "cell_type": "code",
23 | "execution_count": 2,
24 | "metadata": {},
25 | "outputs": [],
26 | "source": [
27 | "working_libs = [\"jupyterlab\"]\n",
28 | "basic_libs = [\"numpy\", \"pandas\"]\n",
29 | "ml_libs = [\"sklearn\", \"joblib\"]\n",
30 | "vis_libs = [\"matplotlib\", \"seaborn\", \"altair\", \"plotnine\", \"modelvis\", \"yellowbrick\"]"
31 | ]
32 | },
33 | {
34 | "cell_type": "code",
35 | "execution_count": 3,
36 | "metadata": {},
37 | "outputs": [],
38 | "source": [
39 | "libs = working_libs + basic_libs + ml_libs + vis_libs"
40 | ]
41 | },
42 | {
43 | "cell_type": "code",
44 | "execution_count": 4,
45 | "metadata": {},
46 | "outputs": [],
47 | "source": [
48 | "def get_version(libs):\n",
49 | " for lib in libs:\n",
50 | " module = importlib.import_module(lib)\n",
51 | " ver = getattr(module, \"__version__\")\n",
52 | " print(ver, \"\\t \", lib)"
53 | ]
54 | },
55 | {
56 | "cell_type": "code",
57 | "execution_count": 5,
58 | "metadata": {},
59 | "outputs": [
60 | {
61 | "name": "stdout",
62 | "output_type": "stream",
63 | "text": [
64 | "0.35.4 \t jupyterlab\n",
65 | "1.15.4 \t numpy\n",
66 | "0.23.4 \t pandas\n",
67 | "0.20.0 \t sklearn\n",
68 | "0.13.0 \t joblib\n",
69 | "3.0.2 \t matplotlib\n",
70 | "0.9.0 \t seaborn\n",
71 | "2.2.2 \t altair\n",
72 | "0.5.1 \t plotnine\n",
73 | "0.1.6 \t modelvis\n"
74 | ]
75 | }
76 | ],
77 | "source": [
78 | "get_version(libs)"
79 | ]
80 | },
81 | {
82 | "cell_type": "code",
83 | "execution_count": 6,
84 | "metadata": {},
85 | "outputs": [],
86 | "source": [
87 | "lib_rqmt = []\n",
88 | "def requirements(libs):\n",
89 | " for lib in libs:\n",
90 | " module = importlib.import_module(lib)\n",
91 | " ver = getattr(module, \"__version__\")\n",
92 | " lib_ver = \"- \"+lib+\">=\"+ver\n",
93 | " lib_rqmt.append(lib_ver)\n",
94 | " print(lib_ver)"
95 | ]
96 | },
97 | {
98 | "cell_type": "code",
99 | "execution_count": 7,
100 | "metadata": {},
101 | "outputs": [
102 | {
103 | "name": "stdout",
104 | "output_type": "stream",
105 | "text": [
106 | "- jupyterlab>=0.35.4\n",
107 | "- numpy>=1.15.4\n",
108 | "- pandas>=0.23.4\n",
109 | "- sklearn>=0.20.0\n",
110 | "- joblib>=0.13.0\n",
111 | "- matplotlib>=3.0.2\n",
112 | "- seaborn>=0.9.0\n",
113 | "- altair>=2.2.2\n",
114 | "- plotnine>=0.5.1\n",
115 | "- modelvis>=0.1.6\n"
116 | ]
117 | }
118 | ],
119 | "source": [
120 | "requirements(libs)"
121 | ]
122 | },
123 | {
124 | "cell_type": "code",
125 | "execution_count": 8,
126 | "metadata": {},
127 | "outputs": [
128 | {
129 | "data": {
130 | "text/plain": [
131 | "['- jupyterlab>=0.35.4',\n",
132 | " '- numpy>=1.15.4',\n",
133 | " '- pandas>=0.23.4',\n",
134 | " '- sklearn>=0.20.0',\n",
135 | " '- joblib>=0.13.0',\n",
136 | " '- matplotlib>=3.0.2',\n",
137 | " '- seaborn>=0.9.0',\n",
138 | " '- altair>=2.2.2',\n",
139 | " '- plotnine>=0.5.1',\n",
140 | " '- modelvis>=0.1.6']"
141 | ]
142 | },
143 | "execution_count": 8,
144 | "metadata": {},
145 | "output_type": "execute_result"
146 | }
147 | ],
148 | "source": [
149 | "lib_rqmt"
150 | ]
151 | },
152 | {
153 | "cell_type": "code",
154 | "execution_count": null,
155 | "metadata": {},
156 | "outputs": [],
157 | "source": []
158 | }
159 | ],
160 | "metadata": {
161 | "kernelspec": {
162 | "display_name": "Python 3",
163 | "language": "python",
164 | "name": "python3"
165 | },
166 | "language_info": {
167 | "codemirror_mode": {
168 | "name": "ipython",
169 | "version": 3
170 | },
171 | "file_extension": ".py",
172 | "mimetype": "text/x-python",
173 | "name": "python",
174 | "nbconvert_exporter": "python",
175 | "pygments_lexer": "ipython3",
176 | "version": "3.7.1"
177 | }
178 | },
179 | "nbformat": 4,
180 | "nbformat_minor": 2
181 | }
182 |
--------------------------------------------------------------------------------
/Module-02b-Explore-Logistic.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "markdown",
5 | "metadata": {},
6 | "source": [
7 | "# Logistic - Explore the Data"
8 | ]
9 | },
10 | {
11 | "cell_type": "markdown",
12 | "metadata": {},
13 | "source": [
14 | "## Raw Data\n",
15 | "\n",
16 | "You are provided with the following data: **loan_data.csv** \n",
17 | "This is the historical data that the bank has provided. It has the following columns\n",
18 | "\n",
19 | "**Application Attributes**:\n",
20 | "- `years`: Number of years the applicant has been employed \n",
21 | "- `ownership`: Whether the applicant owns a house or not \n",
22 | "- `income`: Annual income of the applicant \n",
23 | "- `age`: Age of the applicant \n",
24 | "\n",
25 | "**Behavioural Attributes**:\n",
26 | "- `grade`: Credit grade of the applicant\n",
27 | "\n",
28 | "**Outcome Variable**:\n",
29 | "- `amount` : Amount of Loan provided to the applicant \n",
30 | "- `interest`: Interest rate charged for the applicant \n",
31 | "- `default` : Whether the applicant has defaulted or not "
32 | ]
33 | },
34 | {
35 | "cell_type": "markdown",
36 | "metadata": {},
37 | "source": [
38 | "Let us build some intuition around the Loan Data"
39 | ]
40 | },
41 | {
42 | "cell_type": "markdown",
43 | "metadata": {},
44 | "source": [
45 | "## Frame the Problem\n",
46 | "\n",
47 | "- What are the features\n",
48 | "- What are the target"
49 | ]
50 | },
51 | {
52 | "cell_type": "code",
53 | "execution_count": null,
54 | "metadata": {
55 | "collapsed": true
56 | },
57 | "outputs": [],
58 | "source": []
59 | },
60 | {
61 | "cell_type": "markdown",
62 | "metadata": {},
63 | "source": [
64 | "## Load the Refine Data"
65 | ]
66 | },
67 | {
68 | "cell_type": "code",
69 | "execution_count": 4,
70 | "metadata": {
71 | "collapsed": true
72 | },
73 | "outputs": [],
74 | "source": [
75 | "#Load the libraries\n",
76 | "import numpy as np\n",
77 | "import pandas as pd\n",
78 | "import matplotlib.pyplot as plt\n",
79 | "import seaborn as sns"
80 | ]
81 | },
82 | {
83 | "cell_type": "code",
84 | "execution_count": 5,
85 | "metadata": {
86 | "collapsed": true
87 | },
88 | "outputs": [],
89 | "source": [
90 | "#Default Variables\n",
91 | "%matplotlib inline\n",
92 | "plt.rcParams['figure.figsize'] = (16,9)\n",
93 | "plt.rcParams['font.size'] = 18\n",
94 | "plt.style.use('fivethirtyeight')\n",
95 | "pd.set_option('display.float_format', lambda x: '%.2f' % x)"
96 | ]
97 | },
98 | {
99 | "cell_type": "code",
100 | "execution_count": 6,
101 | "metadata": {
102 | "collapsed": true
103 | },
104 | "outputs": [],
105 | "source": [
106 | "#Load the dataset\n",
107 | "df = pd.read_csv(\"data/loan_data_clean.csv\")"
108 | ]
109 | },
110 | {
111 | "cell_type": "code",
112 | "execution_count": 7,
113 | "metadata": {},
114 | "outputs": [
115 | {
116 | "data": {
117 | "text/html": [
118 | "\n",
119 | "
\n",
120 | " \n",
121 | " \n",
122 | " | \n",
123 | " default | \n",
124 | " amount | \n",
125 | " interest | \n",
126 | " grade | \n",
127 | " years | \n",
128 | " ownership | \n",
129 | " income | \n",
130 | " age | \n",
131 | "
\n",
132 | " \n",
133 | " \n",
134 | " \n",
135 | " | 0 | \n",
136 | " 0 | \n",
137 | " 5000 | \n",
138 | " 10.65 | \n",
139 | " B | \n",
140 | " 10.00 | \n",
141 | " RENT | \n",
142 | " 24000.00 | \n",
143 | " 33 | \n",
144 | "
\n",
145 | " \n",
146 | " | 1 | \n",
147 | " 0 | \n",
148 | " 2400 | \n",
149 | " 10.99 | \n",
150 | " C | \n",
151 | " 25.00 | \n",
152 | " RENT | \n",
153 | " 12252.00 | \n",
154 | " 31 | \n",
155 | "
\n",
156 | " \n",
157 | " | 2 | \n",
158 | " 0 | \n",
159 | " 10000 | \n",
160 | " 13.49 | \n",
161 | " C | \n",
162 | " 13.00 | \n",
163 | " RENT | \n",
164 | " 49200.00 | \n",
165 | " 24 | \n",
166 | "
\n",
167 | " \n",
168 | " | 3 | \n",
169 | " 0 | \n",
170 | " 5000 | \n",
171 | " 10.99 | \n",
172 | " A | \n",
173 | " 3.00 | \n",
174 | " RENT | \n",
175 | " 36000.00 | \n",
176 | " 39 | \n",
177 | "
\n",
178 | " \n",
179 | " | 4 | \n",
180 | " 0 | \n",
181 | " 3000 | \n",
182 | " 10.99 | \n",
183 | " E | \n",
184 | " 9.00 | \n",
185 | " RENT | \n",
186 | " 48000.00 | \n",
187 | " 24 | \n",
188 | "
\n",
189 | " \n",
190 | "
\n",
191 | "
\n",
142 | "\n",
155 | "
\n",
156 | " \n",
157 | " \n",
158 | " | \n",
159 | " default | \n",
160 | " amount | \n",
161 | " grade | \n",
162 | " years | \n",
163 | " ownership | \n",
164 | " income | \n",
165 | " age | \n",
166 | "
\n",
167 | " \n",
168 | " \n",
169 | " \n",
170 | " | 0 | \n",
171 | " 0 | \n",
172 | " 1000 | \n",
173 | " 1 | \n",
174 | " 2.0 | \n",
175 | " 3 | \n",
176 | " 19200.0 | \n",
177 | " 24 | \n",
178 | "
\n",
179 | " \n",
180 | " | 1 | \n",
181 | " 1 | \n",
182 | " 6500 | \n",
183 | " 0 | \n",
184 | " 2.0 | \n",
185 | " 0 | \n",
186 | " 66000.0 | \n",
187 | " 28 | \n",
188 | "
\n",
189 | " \n",
190 | " | 2 | \n",
191 | " 0 | \n",
192 | " 2400 | \n",
193 | " 0 | \n",
194 | " 2.0 | \n",
195 | " 3 | \n",
196 | " 60000.0 | \n",
197 | " 36 | \n",
198 | "
\n",
199 | " \n",
200 | " | 3 | \n",
201 | " 0 | \n",
202 | " 10000 | \n",
203 | " 2 | \n",
204 | " 3.0 | \n",
205 | " 3 | \n",
206 | " 62000.0 | \n",
207 | " 24 | \n",
208 | "
\n",
209 | " \n",
210 | " | 4 | \n",
211 | " 1 | \n",
212 | " 4000 | \n",
213 | " 2 | \n",
214 | " 2.0 | \n",
215 | " 3 | \n",
216 | " 20000.0 | \n",
217 | " 28 | \n",
218 | "
\n",
219 | " \n",
220 | "
\n",
221 | "
\n",
78 | "\n",
91 | "
\n",
92 | " \n",
93 | " \n",
94 | " | \n",
95 | " SalePrice | \n",
96 | " OverallQual | \n",
97 | " GrLivArea | \n",
98 | " GarageCars | \n",
99 | " TotalBsmtSF | \n",
100 | " FullBath | \n",
101 | " YearBuilt | \n",
102 | " FireplaceQu | \n",
103 | " LotFrontage | \n",
104 | "
\n",
105 | " \n",
106 | " \n",
107 | " \n",
108 | " | 0 | \n",
109 | " 208500 | \n",
110 | " 7 | \n",
111 | " 1710 | \n",
112 | " 2 | \n",
113 | " 856 | \n",
114 | " 2 | \n",
115 | " 2003 | \n",
116 | " NaN | \n",
117 | " 65.0 | \n",
118 | "
\n",
119 | " \n",
120 | " | 1 | \n",
121 | " 181500 | \n",
122 | " 6 | \n",
123 | " 1262 | \n",
124 | " 2 | \n",
125 | " 1262 | \n",
126 | " 2 | \n",
127 | " 1976 | \n",
128 | " TA | \n",
129 | " 80.0 | \n",
130 | "
\n",
131 | " \n",
132 | " | 2 | \n",
133 | " 223500 | \n",
134 | " 7 | \n",
135 | " 1786 | \n",
136 | " 2 | \n",
137 | " 920 | \n",
138 | " 2 | \n",
139 | " 2001 | \n",
140 | " TA | \n",
141 | " 68.0 | \n",
142 | "
\n",
143 | " \n",
144 | " | 3 | \n",
145 | " 140000 | \n",
146 | " 7 | \n",
147 | " 1717 | \n",
148 | " 3 | \n",
149 | " 756 | \n",
150 | " 1 | \n",
151 | " 1915 | \n",
152 | " Gd | \n",
153 | " 60.0 | \n",
154 | "
\n",
155 | " \n",
156 | " | 4 | \n",
157 | " 250000 | \n",
158 | " 8 | \n",
159 | " 2198 | \n",
160 | " 3 | \n",
161 | " 1145 | \n",
162 | " 2 | \n",
163 | " 2000 | \n",
164 | " TA | \n",
165 | " 84.0 | \n",
166 | "
\n",
167 | " \n",
168 | "
\n",
169 | "
8 | *Amit Kapoor*
9 | @amitkaps
10 |
11 |
12 | *Bargava Subramanian*
13 | @bargava
14 |
15 | ---
16 |
17 | # **Getting Started**
18 | - Download the Repo: [https://github.com/amitkaps/applied-machine-learning](https://github.com/amitkaps/applied-machine-learning)
19 | - Finish installation
20 | - Run `jupyter notebook` in the console
21 |
22 | ---
23 |
24 | # **Schedule**
25 |
26 | 0900 - 0930: Breakfast
27 | 0930 - 1115: **Session 1** - *Conceptual*
28 | 1115 - 1130: Tea Break
29 | 1130 - 1315: **Session 2** - *Coding*
30 | 1315 - 1400: Lunch
31 | 1400 - 1530: **Session 3** - *Conceptual*
32 | 1530 - 1545: Tea Break
33 | 1545 - 1700: **Session 4** - *Coding*
34 |
35 | ---
36 |
37 | # **Data-Driven Lens**
38 |
39 | > "Data is a clue to the End Truth"
40 | -- Josh Smith
41 |
42 | ---
43 |
44 | # **Metaphor**
45 |
46 | - A start-up providing loans to the consumer
47 | - Running for the last few years
48 | - Now planning to adopt a data-driven lens
49 |
50 | What are the **type of questions** you can ask?
51 |
52 | ---
53 |
54 | # **Type of Questions**
55 | - What is the trend of loan defaults?
56 | - Do older customers have more loan defaults?
57 | - Which customer is likely to have a loan default?
58 | - Why do customers default on their loan?
59 |
60 | ---
61 |
62 | # **Type of Questions**
63 | - Descriptive
64 | - Inquisitive
65 | - Predictive
66 | - Causal
67 |
68 | ---
69 |
70 | # **Data-driven Analytics**
71 | - **Descriptive**: Understand Pattern, Trends, Outlier
72 | - **Inquisitive**: Conduct Hypothesis Testing
73 | - **Predictive**: Make a prediction
74 | - **Causal**: Establish a causal link
75 |
76 | ---
77 |
78 | # **Prediction Challenge**
79 |
80 | > It’s tough to make predictions, especially about the future.
81 | -- Yogi Berra
82 |
83 | ---
84 |
85 | # **How to make a Prediction?**
86 | - **Human Learning**: Make a *Judgement*
87 | - **Machine Programmed**: Create explicit *Rules*
88 | - **Machine Learning**: Learn from *Data*
89 |
90 | ---
91 |
92 | # **Machine Learning (ML)**
93 |
94 | > [Machine learning is the] field of study that gives computers the ability to learn without being explicitly programmed.
95 | -- *Arthur Samuel*
96 |
97 |
98 |
99 | > Machine learning is the study of computer algorithm that improve automatically through experience
100 | -- *Tom Mitchell*
101 |
102 | ---
103 |
104 | # **Machine Learning: Essense**
105 |
106 | - A pattern exists
107 | - It cannot be pinned down mathematically
108 | - Have data on it to learn from
109 |
110 | > "Use a set of observations (data) to uncover an underlying process"
111 |
112 | ---
113 |
114 | # **Machine Learning**
115 | - **Theory**
116 | - **Paradigms**
117 | - **Models**
118 | - **Methods**
119 | - **Process**
120 |
121 | ---
122 |
123 | # **Applied ML - Approach**
124 | - **Theory**: Understand Key Concepts (Intuition)
125 | - **Paradigms**: Limit to One (Supervised)
126 | - **Models**: Use Two Types (Linear, Trees)
127 | - **Methods**: Apply Key Ones (Validation, Selection)
128 | - **Process**: Code the Approach (Real Examples)
129 |
130 | ---
131 |
132 | # **ML Theory: Data Types**
133 |
134 | - What are the types of data on which we are learning?
135 | - Can you give example of say measuring temperature?
136 |
137 | ---
138 |
139 | # **Data Types e.g. Temperature**
140 |
141 | - **Categorical**
142 | - *Nominal*: Burned, Not Burned
143 | - *Ordinal*: Hot, Warm, Cold
144 | - **Continuous**
145 | - *Interval*: 30 °C, 40 °C, 80 °C
146 | - *Ratio*: 30 K, 40 K, 50 K
147 |
148 | ---
149 |
150 | # **Data Types - Operations**
151 |
152 | - **Categorical**
153 | - *Nominal*: = , !=
154 | - *Ordinal*: =, !=, >, <
155 | - **Continuous**
156 | - *Interval*: =, !=, >, <, -, % of diff
157 | - *Ratio*: =, !=, >, <, -, +, %
158 |
159 | ---
160 |
161 | # **Case Example**
162 |
163 | *Context*: Loan Approval
164 |
165 | *Customer Application*
166 | - **age**: age of the applicant
167 | - **income**: annual income of the applicant
168 | - **year**: no. of years of employment
169 | - **ownership**: type of house owned
170 | - **grade**: credit grade for the applicant
171 |
172 | *Question* - How much loan **amount** to approve?
173 |
174 | ---
175 |
176 | # **Historical Data**
177 |
178 | ```
179 | age income years ownership grade amount
180 | --- ------- ----- --------- ------- -------
181 | 31 12252 25.0 RENT C 2400
182 | 24 49200 13.0 RENT C 10000
183 | 28 75000 11.0 OWN B 12000
184 | 27 110000 13.0 MORTGAGE A 3600
185 | 33 24000 10.0 RENT B 5000
186 |
187 | ```
188 |
189 | ---
190 |
191 | # **Data Types**
192 |
193 | - **Categorical**
194 | - *Nominal*: home owner [rent, own, mortgage]
195 | - *Ordinal*: credit grade [A > B > C > D > E]
196 | - **Continuous**
197 | - *Interval*: approval date [20/04/16, 19/11/15]
198 | - *Ratio*: loan amount [3000, 10000]
199 |
200 | ---
201 |
202 | # **ML Terminology**
203 |
204 | **Features**: $$\mathbf{x}$$
205 | - `age`, `income`, `years`, `ownership`, `grade`
206 |
207 |
208 | **Target**: $$y$$
209 | - `amount`
210 |
211 | **Training Data**: $$ (\mathbf{x}_{1}, y_{1}), (\mathbf{x}_{2}, y_{2}) ... (\mathbf{x}_{n}, y_{n}) $$
212 | - historical records
213 |
214 | ---
215 |
216 | # **ML Paradigm: Supervised**
217 |
218 | Given a set of **feature** $$\mathbf{x}$$, to predict the value of **target** $$y$$
219 |
220 | Learning Paradigm: **Supervised**
221 |
222 | - If $$y$$ is *continuous* - **Regression**
223 | - If $$y$$ is *categorical* - **Classification**
224 |
225 | ---
226 |
227 | # **ML Theory: Formulation**
228 | - **Features** $$\mathbf{x}$$ *(customer application)*
229 | - **Target** $$y$$ *(loan amount)*
230 | - **Target Function** $$\mathcal{f}: \mathcal{X} \to \mathcal{y}$$ (ideal formula)
231 | - **Data** $$ (\mathbf{x}_{1}, y_{1}), (\mathbf{x}_{2}, y_{2}) ... (\mathbf{x}_{n}, y_{n}) $$ *(historical records)*
232 | - **Final Hypothesis** $$\mathcal{g}: \mathcal{X} \to \mathcal{y}$$ (formula to use)
233 | - **Hypothesis Set** $$ \mathcal{H} $$ (all possible formulas)
234 | - **Learning Algorithm** $$\mathcal{A}$$ (how to learn the formula)
235 |
236 |
237 | ---
238 |
239 | # **ML Theory: Formulation**
240 |
241 | $$\text{unknown target function}$$
242 | $$\mathcal{f}: \mathcal{X} \to \mathcal{y}$$
243 | $$ | $$
244 | $$\text{training data}$$
245 | $$ (\mathbf{x}_{1}, y_{1}), (\mathbf{x}_{2}, y_{2}) ... (\mathbf{x}_{n}, y_{n}) $$
246 | $$ | $$
247 | $$\text{hypothesis set} \quad \rightarrow \quad \text{learning algorithm} \qquad \qquad \qquad \qquad $$
248 | $$ \mathcal{H} \qquad \qquad \qquad \qquad \qquad \mathcal{A} \qquad \qquad \qquad \qquad \qquad $$
249 | $$ | $$
250 | $$\text{final hypothesis}$$
251 | $$\mathcal{g} \to \mathcal{f}$$
252 |
253 | ---
254 |
255 | # **ML Theory: Learning Model**
256 |
257 | The Learning Model is composed of the two elements
258 |
259 | - The Hypothesis Set: $$ \mathcal{H} = \{\mathcal{h}\} \qquad \mathcal{g} \in \mathcal{H} $$
260 | - Learning Algorithm: $$ \mathcal{A} $$
261 |
262 | ---
263 |
264 | # **ML Theory: Formulation (Simplified)**
265 |
266 | $$\text{unknown target function}$$
267 | $$ y = \mathcal{f}(\mathbf{x})$$
268 | $$ | $$
269 | $$\text{training data}$$
270 | $$ (\mathbf{x}_{1}, y_{1}), (\mathbf{x}_{2}, y_{2}) ... (\mathbf{x}_{n}, y_{n}) $$
271 | $$ | $$
272 | $$\text{hypothesis set} \quad \rightarrow \quad \text{learning algorithm} \qquad \qquad \qquad \qquad $$
273 | $$ \{ \mathcal{h}(\mathbf{x})\} \qquad \qquad \qquad \qquad \mathcal{A} \qquad \qquad \qquad \qquad \qquad $$
274 | $$ | $$
275 | $$\text{final hypothesis}$$
276 | $$\mathcal{g}(\mathbf{x}) \to \mathcal{f}(\mathbf{x})$$
277 |
278 | ---
279 |
280 | # **Linear Algorithms**
281 |
282 | $$ s = \sum_{i=1}^d w_{i} x_{i} $$
283 |
284 |
285 | 
286 |
287 | ---
288 |
289 | # **Simple Hypothesis Set: Linear Regression**
290 |
291 | For $$d$$ features in training data,
292 |
293 | $$ h(\mathbf{x}) = \sum_{i=1}^d w_{i} x_{i} $$
294 |
295 | How do we choose the right $$w_{i}$$?
296 |
297 | ---
298 |
299 | # **Error**
300 |
301 | 
302 |
303 | [^1]: Source - Learning from Data
304 |
305 | ---
306 |
307 | # **Error Measure - MSE**
308 |
309 | How well does $$ h(\mathbf{x}) $$ approximate to $$ f(\mathbf{x}) $$
310 |
311 | We will use squared error $$ {( h(\mathbf{x}) - f(\mathbf{x}))}^2 $$
312 |
313 | $$ E_{in}(h) = \frac{1}{N} \sum_{i=i}^N {( h(\mathbf{x}) - y_i))}^2 $$
314 |
315 | ---
316 |
317 | # **Learning Algorithm - Linear Regression**
318 |
319 | - Linear Regression algorithm aims to minimise $$ E_{in}(h)$$
320 | - **One-Step Learning** -> Solves to give $$g(\mathbf{x})$$
321 |
322 | $$g(\mathbf{x}) = \hat{y} $$
323 |
324 |
325 | $$ E_{in}(g) = \frac{1}{N} \sum_{i=1}^N {( \hat{y}_i - y_i)}^2 $$
326 |
327 |
328 | ---
329 |
330 | # **Machine Learning Process**
331 |
332 | - *Frame*: Problem definition
333 | - *Acquire*: Data ingestion
334 | - *Refine*: Data wrangling
335 | - *Transform*: Feature creation
336 | - *Explore*: Feature selection
337 | - *Model*: Model creation & assessment
338 | - *Insight*: Communication
339 |
340 | ---
341 |
342 | # **Frame**
343 |
344 | **Variables**
345 | - `age`, `income`, `years`, `ownership`, `grade`, `amount`, `default` and `interest`
346 |
347 | - What are the **Features**: $$\mathbf{x}$$ ?
348 | - What are the **Target**: $$y$$
349 |
350 | ---
351 |
352 | # **Frame**
353 |
354 | **Features**: $$\mathbf{x}$$
355 | - `age`
356 | - `income`
357 | - `years`,
358 | - `ownership`
359 | - `grade`,
360 |
361 | **Target**: $$y$$
362 | - `amount` * (1 - `default`)
363 |
364 |
365 | ---
366 |
367 | # **Acquire**
368 |
369 | - Simple! Just read the data from `csv` file
370 |
371 | ---
372 |
373 | # **Refine - Missing Value**
374 |
375 | - **REMOVE** - NAN rows
376 | - **IMPUTATION** - Replace them with something?
377 | - Mean
378 | - Median
379 | - Fixed Number - Domain Relevant
380 | - High Number (999) - Issue with modelling
381 | - **BINNING** - Categorical variable and "Missing becomes a category*
382 | - **DOMAIN SPECIFIC** - Entry error, pipeline, etc.
383 |
384 | ---
385 |
386 | # **Refine - Outlier Treatment**
387 |
388 | - What is an outlier?
389 | - Descriptive Plots
390 | - Histogram
391 | - Box-Plot
392 | - Measuring
393 | - Z-score
394 | - Modified Z-score > 3.5
395 | where modified Z-score = 0.6745 * (x - x_median) / MAD
396 |
397 | ---
398 |
399 | # **Explore**
400 |
401 | - Single Variable Exploration
402 | - Dual Variable Exploration
403 | - Multi Variable Exploration
404 |
405 | ---
406 |
407 | # **Transform**
408 |
409 | Encodings
410 | - One Hot Encoding
411 | - Label Encoding
412 |
413 | Feature Transformation
414 | - Log Transform
415 | - Sqrt Transform
416 |
417 | ---
418 |
419 | # **Model - Linear Regression**
420 |
421 | **Parameters**
422 | - fit_intercept
423 | - normalization
424 |
425 | **Error Measure**
426 | - mean squared error
427 |
428 | ---
429 |
430 |
431 | # **Real-World Challenge - Noise**
432 |
433 | - The "target function" $$f$$ is not always a *function*
434 | - Not unique target value for same input
435 | - Need to add noise $$N(0,\sigma)$$
436 |
437 | $$ y = f(\mathbf{x}) + \epsilon(\mathbf{x}) $$
438 |
439 | ---
440 |
441 | # **Noise Implication**
442 |
443 | The best model we can create will have an expected error of $$\sigma^2$$
444 |
445 | If Noise ($$\sigma$$) is large, that means feature set does not capture large enough factors in the underlying process
446 | - Need to create **better features**
447 | - Need to find **new features**
448 |
449 | ---
450 |
451 | # **When are we learning?**
452 |
453 | Learning is defined as $$g≈f$$, which happens when
454 |
455 | (1) Can we make $$E_{out}(g)$$ is close enough to $$E_{in}(g)$$?
456 |
457 | $$E_{out}(g)≈E_{in}(g)$$
458 |
459 | (1) Can we make $$E_{in}(g)$$ small enough?
460 |
461 | $$E_{in}(g)≈0$$
462 |
463 | ---
464 |
465 | # **ML Theory: Generalisation**
466 |
467 | 
468 |
469 | For Learning, $$E_{out}(g)≈E_{in}(g)$$
470 |
471 | To find the generalisation error, we need to split our data into training and test samples
472 |
473 | Given large $$N$$, the expected generalisation error should be zero
474 |
475 | ---
476 |
477 | # **ML Theory: Generalisation**
478 |
479 | For Learning, $$E_{in}(g)≈0$$
480 |
481 | **Complex Model**: Better chance of approximating $$f$$
482 | **Simple Model**: Better chance of generalising $$E_{out}$$
483 |
484 | Lets try by increasing the model complexity - More features through interaction effect
485 |
486 | ---
487 |
488 | # **ML Theory: Model Complexity**
489 |
490 | 
491 |
492 | ---
493 |
494 | 
495 | # **ML Theory: Bias-Variance**
496 |
497 | For Learning, $$E_{in}(g)≈0$$
498 |
499 | Given large $$N$$, the expected error should be the bias
500 |
501 | - **Bias** are the simplifying assumptions made by a model to make the target function easier to learn.
502 | - **Variance** is the amount that the estimate of the target function will change if different training data was used.
503 |
504 | ---
505 |
506 | # **ML Theory: Bias-Variance Tradeoff**
507 |
508 | 
509 |
510 | ---
511 |
512 | 
513 |
514 | # **ML Theory: Overfitting**
515 |
516 | - Simple Target Function
517 | - 5th data point - noisy
518 | - 4th order polynomial fit
519 |
520 | $$E_{in}=0$$, $$E_{out}$$ is large
521 |
522 | *Overfitting* - Fitting the data more than warranted, and hence **fitting the noise**
523 |
524 | ---
525 |
526 | # **ML Theory: Addressing Overfitting**
527 |
528 | $$E_{out}(h) = E_{in}(h) + \text{overfit penalty}$$
529 |
530 | - **Regularization**: Not letting the weights grow
531 | - Ridge: add $$||w||^2$$ to error minimisation
532 | - Lasso: add $$||w||$$ to error minimisation
533 | - **Validation**: Checking when we reach bottom point
534 |
535 | ---
536 |
537 | # **Regularization - Ridge**
538 |
539 | $$ Minimize \quad E_{in}(w) + \frac{\lambda}{N}||w||^2 $$
540 |
541 | 
542 |
543 | ---
544 |
545 | 
546 |
547 | # **Validation**
548 |
549 | Validation set: $$K$$
550 | Training set: $$N-K$$
551 |
552 | Rule of Thumb: $$N = \frac{K}{5}$$
553 |
554 | Note: The validation set is used for learning
555 |
556 | ---
557 |
558 | # **Cross Validation**
559 |
560 | Repeats the process 5-times
561 |
562 | 
563 |
564 | ---
565 |
566 | 
567 |
568 | # **Model Selection**
569 |
570 | How to choose between competing model?
571 |
572 | Choose the function $$g_{m}$$ with
573 | lowest cross-validation error $$E_{m}$$
574 |
575 | ---
576 |
577 | # **Applied ML**
578 | - **Theory**: Formulation, Generalisation, Bias-Variance, Overfitting
579 | - **Paradigms**: Supervised - Regression
580 | - **Models**: Linear - OLS, Ridge, Lasso
581 | - **Methods**: Regularisation, Validation
582 | - **Process**: Frame, Acquire, Refine, Transform, Explore, Model
583 |
584 |
585 | ---
586 |
587 | ## **Classification Problem**
588 |
589 | *Context*: Loan Default
590 |
591 | *Customer Application*
592 | - **age**: age of the applicant
593 | - **income**: annual income of the applicant
594 | - **year**: no. of years of employment
595 | - **ownership**: type of house owned
596 | - **grade**: credit grade for the applicant
597 | - **amount**: loan amount given
598 | - **interest**: interest rate of loan
599 |
600 | *Question* - Who is likely to **default**?
601 |
602 | ---
603 |
604 | # **Linear Models**
605 |
606 | $$ s = \sum_{i=1}^d w_{i} x_{i} $$
607 |
608 |
609 | 
610 |
611 | ---
612 |
613 | # **Logit Function**
614 |
615 | ### $$ \theta (s)={\frac {e^{s}}{e^{s}+1}}={\frac {1}{1+e^{-s}}}$$
616 |
617 | 
618 |
619 | ---
620 |
621 | # **Logistic Relationship**
622 |
623 | Find the $$ w_{i} $$ weights that best fit:
624 | $$ y=1 $$ if $$ \sum_{i=1}^d w_{i} x_{i} > 0$$
625 | $$ y=0$$, otherwise
626 |
627 | Follows:
628 |
629 | $$ \theta(y_i)={\frac {1}{1+e^{-(\sum_{i=1}^d w_{i} x_{i})}}} $$
630 |
631 | ---
632 |
633 | # **Error - Likelihood / Probabilities**
634 |
635 | Where, $$h(\mathbf{x}) = \sum_{i=1}^d w_{i} x_{i} $$
636 |
637 | Minimise the **log-likelihood** values
638 |
639 | $$E(\mathbf{h}) = - \frac{1}{N} ln \left( \prod_{i=1}^N \theta (y_i h(\mathbf{x})) \right)$$
640 |
641 |
642 |
643 | ---
644 |
645 | # **Learning Algorithm - Logistic**
646 |
647 | - Logistic Regression algorithm aims to minimise $$ E_{in}(h)$$
648 | - **Iterative Method** -> Solves to give $$g(\mathbf{x})$$
649 |
650 | $$g(\mathbf{x}) = \hat{y} $$
651 |
652 | $$ E_{in}(g) = \frac{1}{N} \sum_{i=1}^N ln( 1 + e^{-y_i \hat{y_i}})$$
653 |
654 |
655 | ---
656 |
657 | # **Error Metric - Confusion Matrix**
658 |
659 | 
660 |
661 | ---
662 |
663 | # **Model Evaluation**
664 |
665 | **Classification Metrics**
666 |
667 | 
668 |
669 | Recall (TPR) = TP / (TP + FN)
670 |
671 | Precision = TP / (TP + FP)
672 |
673 | Specificity (TNR) = TN / (TN + FP)
674 |
675 | ---
676 |
677 | # **Model Evaluation**
678 |
679 | **Receiver Operating Characteristic Curve**
680 |
681 | Plot of TPR vs FPR at different discrimination threshold
682 |
683 | 
684 |
685 | ---
686 |
687 | # **Decision Tree**
688 |
689 | Example: Survivor on Titanic
690 |
691 | 
692 |
693 | ---
694 |
695 | # **Decision Tree**
696 |
697 | - Easy to interpret
698 | - Little data preparation
699 | - Scales well with data
700 | - White-box model
701 | - Instability – changing variables, altering sequence
702 | - Overfitting
703 |
704 | ---
705 |
706 | # **Bagging**
707 |
708 | - Also called bootstrap aggregation, reduces variance
709 | - Uses decision trees and uses a model averaging approach
710 |
711 | ---
712 |
713 | # **Random Forest**
714 |
715 | - Combines bagging idea and random selection of features.
716 | - Similar to decision trees are constructed – but at each split, a random subset of features is used.
717 |
718 | 
719 |
720 | ---
721 |
722 | > If you torture the data enough, it will confess.
723 | -- Ronald Case
724 |
725 | ---
726 |
727 | # **Challenges**
728 | - Data Snooping
729 | - Selection Bias
730 | - Survivor Bias
731 | - Omitted Variable Bias
732 | - Black-box model Vs White-Box model
733 | - Adherence to regulations
734 |
735 | ---
736 |
737 |
738 | # **Applied ML**
739 | - **Theory**: Formulation, Generalisation, Bias-Variance, Overfitting
740 | - **Paradigms**: Supervised - Regression & Classification
741 | - **Models**: Linear Models, Tree Models
742 | - **Methods**: Regularisation, Validation, Aggregation
743 | - **Process**: Frame, Acquire, Refine, Transform, Explore, Model
744 |
745 |
746 | ---
--------------------------------------------------------------------------------
/outline.pdf:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/amitkaps/applied-machine-learning/9752d1cd19a530b36c81632c291a6607ab3049d3/outline.pdf
--------------------------------------------------------------------------------
/pre-requisites.md:
--------------------------------------------------------------------------------
1 | # Pre-requisites
2 |
3 | - Programming knowledge is mandatory. Attendee should be able to write conditional statements, use loops, be comfortable writing functions and be able to understand code snippets and come up with programming logic.
4 | - Participants should have a basic familiarity of Python. Specifically, we expect participants to know the first three sections from this: [http://anandology.com/python-practice-book/](http://anandology.com/python-practice-book/)
5 | - Participants should have experience with using `Pandas` and `Jupyter Notebook`. At the bare minimum, you should be able to understand and run the code in this [The Art of Data Science](https://github.com/amitkaps/art-data-science) repo. Refer to the Onion Notebook's and especially the Acquire, Refine, Transform and Explore sections.
6 |
--------------------------------------------------------------------------------
/reference/.Module-01b-reference.ipynb.icloud:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/amitkaps/applied-machine-learning/9752d1cd19a530b36c81632c291a6607ab3049d3/reference/.Module-01b-reference.ipynb.icloud
--------------------------------------------------------------------------------
/reference/Module-01c-reference.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "markdown",
5 | "metadata": {},
6 | "source": [
7 | "# Transform and Model"
8 | ]
9 | },
10 | {
11 | "cell_type": "markdown",
12 | "metadata": {},
13 | "source": [
14 | "Let us build a Regression Model for prediciting the amount to be approved"
15 | ]
16 | },
17 | {
18 | "cell_type": "code",
19 | "execution_count": 2,
20 | "metadata": {
21 | "collapsed": true
22 | },
23 | "outputs": [],
24 | "source": [
25 | "#Load the libraries\n",
26 | "import numpy as np\n",
27 | "import pandas as pd\n",
28 | "import matplotlib.pyplot as plt\n",
29 | "import seaborn as sns"
30 | ]
31 | },
32 | {
33 | "cell_type": "code",
34 | "execution_count": 3,
35 | "metadata": {
36 | "collapsed": true
37 | },
38 | "outputs": [],
39 | "source": [
40 | "#Default Variables\n",
41 | "%matplotlib inline\n",
42 | "plt.rcParams['figure.figsize'] = (16,9)\n",
43 | "plt.rcParams['font.size'] = 18\n",
44 | "plt.style.use('fivethirtyeight')\n",
45 | "pd.set_option('display.float_format', lambda x: '%.2f' % x)"
46 | ]
47 | },
48 | {
49 | "cell_type": "code",
50 | "execution_count": 4,
51 | "metadata": {
52 | "collapsed": true
53 | },
54 | "outputs": [],
55 | "source": [
56 | "#Load the dataset\n",
57 | "df = pd.read_csv(\"data/loan_data_clean.csv\")"
58 | ]
59 | },
60 | {
61 | "cell_type": "code",
62 | "execution_count": 5,
63 | "metadata": {
64 | "collapsed": false
65 | },
66 | "outputs": [
67 | {
68 | "data": {
69 | "text/html": [
70 | "\n",
71 | "
\n",
72 | " \n",
73 | " \n",
74 | " | \n",
75 | " default | \n",
76 | " amount | \n",
77 | " interest | \n",
78 | " grade | \n",
79 | " years | \n",
80 | " ownership | \n",
81 | " income | \n",
82 | " age | \n",
83 | "
\n",
84 | " \n",
85 | " \n",
86 | " \n",
87 | " | 0 | \n",
88 | " 0 | \n",
89 | " 5000 | \n",
90 | " 10.65 | \n",
91 | " B | \n",
92 | " 10.00 | \n",
93 | " RENT | \n",
94 | " 24000.00 | \n",
95 | " 33 | \n",
96 | "
\n",
97 | " \n",
98 | " | 1 | \n",
99 | " 0 | \n",
100 | " 2400 | \n",
101 | " 10.99 | \n",
102 | " C | \n",
103 | " 25.00 | \n",
104 | " RENT | \n",
105 | " 12252.00 | \n",
106 | " 31 | \n",
107 | "
\n",
108 | " \n",
109 | " | 2 | \n",
110 | " 0 | \n",
111 | " 10000 | \n",
112 | " 13.49 | \n",
113 | " C | \n",
114 | " 13.00 | \n",
115 | " RENT | \n",
116 | " 49200.00 | \n",
117 | " 24 | \n",
118 | "
\n",
119 | " \n",
120 | " | 3 | \n",
121 | " 0 | \n",
122 | " 5000 | \n",
123 | " 10.99 | \n",
124 | " A | \n",
125 | " 3.00 | \n",
126 | " RENT | \n",
127 | " 36000.00 | \n",
128 | " 39 | \n",
129 | "
\n",
130 | " \n",
131 | " | 4 | \n",
132 | " 0 | \n",
133 | " 3000 | \n",
134 | " 10.99 | \n",
135 | " E | \n",
136 | " 9.00 | \n",
137 | " RENT | \n",
138 | " 48000.00 | \n",
139 | " 24 | \n",
140 | "
\n",
141 | " \n",
142 | "
\n",
143 | "
samples = 18
value = [4, 6, 8]
class = High>, fillcolor="#8139e52a"] ;
5 | 1 [label=samples = 6
value = [0, 0, 6]
class = High>, fillcolor="#8139e5ff"] ;
6 | 0 -> 1 [labeldistance=2.5, labelangle=45, headlabel="True"] ;
7 | 2 [label=gini = 0.611
samples = 12
value = [4, 6, 2]
class = Moderate>, fillcolor="#39e58140"] ;
8 | 0 -> 2 [labeldistance=2.5, labelangle=-45, headlabel="False"] ;
9 | 3 [label=gini = 0.444
samples = 6
value = [0, 4, 2]
class = Moderate>, fillcolor="#39e5817f"] ;
10 | 2 -> 3 ;
11 | 4 [label=samples = 2
value = [0, 0, 2]
class = High>, fillcolor="#8139e5ff"] ;
12 | 3 -> 4 ;
13 | 5 [label=samples = 4
value = [0, 4, 0]
class = Moderate>, fillcolor="#39e581ff"] ;
14 | 3 -> 5 ;
15 | 6 [label=gini = 0.444
samples = 6
value = [4, 2, 0]
class = Low>, fillcolor="#e581397f"] ;
16 | 2 -> 6 ;
17 | 7 [label=samples = 2
value = [0, 2, 0]
class = Moderate>, fillcolor="#39e581ff"] ;
18 | 6 -> 7 ;
19 | 8 [label=samples = 4
value = [4, 0, 0]
class = Low>, fillcolor="#e58139ff"] ;
20 | 6 -> 8 ;
21 | }
--------------------------------------------------------------------------------
/tree2.dot:
--------------------------------------------------------------------------------
1 | digraph Tree {
2 | node [shape=box, style="filled, rounded", color="black", fontname=helvetica] ;
3 | edge [fontname=helvetica] ;
4 | 0 [label=gini = 0.499
samples = 7727
value = [4030, 3697]
class = no>, fillcolor="#e5813915"] ;
5 | 1 [label=gini = 0.466
samples = 4555
value = [2869, 1686]
class = no>, fillcolor="#e5813969"] ;
6 | 0 -> 1 [labeldistance=2.5, labelangle=45, headlabel="True"] ;
7 | 2 [label=gini = 0.411
samples = 2099
value = [1492, 607]
class = no>, fillcolor="#e5813997"] ;
8 | 1 -> 2 ;
9 | 3 [label=gini = 0.498
samples = 200
value = [107, 93]
class = no>, fillcolor="#e5813921"] ;
10 | 2 -> 3 ;
11 | 4 [label=samples = 149
value = [84, 65]
class = no>, fillcolor="#e581393a"] ;
12 | 3 -> 4 ;
13 | 5 [label=samples = 51
value = [23, 28]
class = yes>, fillcolor="#399de52e"] ;
14 | 3 -> 5 ;
15 | 6 [label=gini = 0.395
samples = 1899
value = [1385, 514]
class = no>, fillcolor="#e58139a0"] ;
16 | 2 -> 6 ;
17 | 7 [label=samples = 1815
value = [1313, 502]
class = no>, fillcolor="#e581399e"] ;
18 | 6 -> 7 ;
19 | 8 [label=samples = 84
value = [72, 12]
class = no>, fillcolor="#e58139d4"] ;
20 | 6 -> 8 ;
21 | 9 [label=gini = 0.493
samples = 2456
value = [1377, 1079]
class = no>, fillcolor="#e5813937"] ;
22 | 1 -> 9 ;
23 | 10 [label=gini = 0.497
samples = 1587
value = [856, 731]
class = no>, fillcolor="#e5813925"] ;
24 | 9 -> 10 ;
25 | 11 [label=samples = 1585
value = [856, 729]
class = no>, fillcolor="#e5813926"] ;
26 | 10 -> 11 ;
27 | 12 [label=samples = 2
value = [0, 2]
class = yes>, fillcolor="#399de5ff"] ;
28 | 10 -> 12 ;
29 | 13 [label=gini = 0.48
samples = 869
value = [521, 348]
class = no>, fillcolor="#e5813955"] ;
30 | 9 -> 13 ;
31 | 14 [label=samples = 281
value = [188, 93]
class = no>, fillcolor="#e5813981"] ;
32 | 13 -> 14 ;
33 | 15 [label=samples = 588
value = [333, 255]
class = no>, fillcolor="#e581393c"] ;
34 | 13 -> 15 ;
35 | 16 [label=gini = 0.464
samples = 3172
value = [1161, 2011]
class = yes>, fillcolor="#399de56c"] ;
36 | 0 -> 16 [labeldistance=2.5, labelangle=-45, headlabel="False"] ;
37 | 17 [label=gini = 0.468
samples = 3060
value = [1144, 1916]
class = yes>, fillcolor="#399de567"] ;
38 | 16 -> 17 ;
39 | 18 [label=gini = 0.481
samples = 1685
value = [677, 1008]
class = yes>, fillcolor="#399de554"] ;
40 | 17 -> 18 ;
41 | 19 [label=samples = 1650
value = [669, 981]
class = yes>, fillcolor="#399de551"] ;
42 | 18 -> 19 ;
43 | 20 [label=samples = 35
value = [8, 27]
class = yes>, fillcolor="#399de5b3"] ;
44 | 18 -> 20 ;
45 | 21 [label=gini = 0.449
samples = 1375
value = [467, 908]
class = yes>, fillcolor="#399de57c"] ;
46 | 17 -> 21 ;
47 | 22 [label=samples = 1346
value = [450, 896]
class = yes>, fillcolor="#399de57f"] ;
48 | 21 -> 22 ;
49 | 23 [label=samples = 29
value = [17, 12]
class = no>, fillcolor="#e581394b"] ;
50 | 21 -> 23 ;
51 | 24 [label=gini = 0.257
samples = 112
value = [17, 95]
class = yes>, fillcolor="#399de5d1"] ;
52 | 16 -> 24 ;
53 | 25 [label=samples = 1
value = [1, 0]
class = no>, fillcolor="#e58139ff"] ;
54 | 24 -> 25 ;
55 | 26 [label=gini = 0.247
samples = 111
value = [16, 95]
class = yes>, fillcolor="#399de5d4"] ;
56 | 24 -> 26 ;
57 | 27 [label=samples = 6
value = [2, 4]
class = yes>, fillcolor="#399de57f"] ;
58 | 26 -> 27 ;
59 | 28 [label=samples = 105
value = [14, 91]
class = yes>, fillcolor="#399de5d8"] ;
60 | 26 -> 28 ;
61 | }
--------------------------------------------------------------------------------
/tree_3.dot:
--------------------------------------------------------------------------------
1 | digraph Tree {
2 | node [shape=box, style="filled, rounded", color="black", fontname=helvetica] ;
3 | edge [fontname=helvetica] ;
4 | 0 [label=gini = 0.499
samples = 7727
value = [4030, 3697]
class = no>, fillcolor="#e5813915"] ;
5 | 1 [label=gini = 0.498
samples = 7316
value = [3860, 3456]
class = no>, fillcolor="#e581391b"] ;
6 | 0 -> 1 [labeldistance=2.5, labelangle=45, headlabel="True"] ;
7 | 2 [label=gini = 0.5
samples = 2411
value = [1186, 1225]
class = yes>, fillcolor="#399de508"] ;
8 | 1 -> 2 ;
9 | 3 [label=samples = 975
value = [437, 538]
class = yes>, fillcolor="#399de530"] ;
10 | 2 -> 3 ;
11 | 4 [label=samples = 1436
value = [749, 687]
class = no>, fillcolor="#e5813915"] ;
12 | 2 -> 4 ;
13 | 5 [label=gini = 0.496
samples = 4905
value = [2674, 2231]
class = no>, fillcolor="#e581392a"] ;
14 | 1 -> 5 ;
15 | 6 [label=samples = 772
value = [386, 386]
class = no>, fillcolor="#e5813900"] ;
16 | 5 -> 6 ;
17 | 7 [label=samples = 4133
value = [2288, 1845]
class = no>, fillcolor="#e5813931"] ;
18 | 5 -> 7 ;
19 | 8 [label=gini = 0.485
samples = 411
value = [170, 241]
class = yes>, fillcolor="#399de54b"] ;
20 | 0 -> 8 [labeldistance=2.5, labelangle=-45, headlabel="False"] ;
21 | 9 [label=gini = 0.471
samples = 340
value = [129, 211]
class = yes>, fillcolor="#399de563"] ;
22 | 8 -> 9 ;
23 | 10 [label=samples = 35
value = [19, 16]
class = no>, fillcolor="#e5813928"] ;
24 | 9 -> 10 ;
25 | 11 [label=samples = 305
value = [110, 195]
class = yes>, fillcolor="#399de56f"] ;
26 | 9 -> 11 ;
27 | 12 [label=gini = 0.488
samples = 71
value = [41, 30]
class = no>, fillcolor="#e5813944"] ;
28 | 8 -> 12 ;
29 | 13 [label=samples = 68
value = [41, 27]
class = no>, fillcolor="#e5813957"] ;
30 | 12 -> 13 ;
31 | 14 [label=samples = 3
value = [0, 3]
class = yes>, fillcolor="#399de5ff"] ;
32 | 12 -> 14 ;
33 | }
--------------------------------------------------------------------------------