Table of Contents
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
1. Data Analytics in the Modern Enterprise. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
The Evolution of Data Analytics 1
Different Types of Data Analytics 7
Descriptive Analytics 8
Diagnostic Analytics 9
Predictive Analytics 10
Prescriptive Analytics 11
Knowledge Acquisition, Machine Learning, and
the Role of Predictive Analytics 13
Tools, Frameworks, and Platforms in the Predictive Analytics World 16
Languages and Libraries 17
Services 18
Conclusion 21
2. Predictive Analytics: An Operational Necessity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
The Move from “Data Producing” to “Data Driven” 23
Challenges to Using Predictive Analytics 26
People 27
Data 28
Technology 29
Vertical Industry Use Cases for Predictive Analytics 31
Finance 31
Healthcare 34
Automotive 36
Entertainment 38
Conclusion 44
v
3. The Mathematics and Algorithms Behind Predictive Analytics. . . . . . . . . . . . . . . . . . . . . 45
Statistics and Linear Algebra 46
Regression 53
What Is Regression Analysis? 56
Regression Techniques 59
R-squared and P-value 67
Selecting a Regression Model 70
Decision Trees 72
Training Decision Trees 75
Using Decision Trees to Solve Regression Problems: Regression Trees 80
Tuning Decision Trees 82
Other Algorithms 84
Random Forests 84
Neural Networks 86
Support Vector Machines 93
Naive Bayes Classifier 98
Other Learning Patterns in Machine Learning 105
Conclusion 109
4. Working with Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Understanding Data 111
Data Preprocessing and Feature Engineering 114
Handling Missing Data 114
Categorical Data Encoding 115
Data Transformation 116
Outlier Management 116
Handling Imbalanced Data 117
Combining Data 118
Feature Selection 119
Splitting Preprocessed Data 120
Understanding Bias 120
The Predictive Analytics Pipeline 122
The Data Stage 122
The Model Stage 123
The Serving Stage 124
Other Components 126
Selecting the Right Model 127
Conclusion 128
5. Python and scikit-learn for Predictive Analytics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Anaconda and Jupyter Notebooks 129
NumPy in Python 132
vi | Table of Contents
Introduction to NumPy 133
Generating Arrays 137
Array Slicing 139
Array Transformation 140
Other Array Operations 141
Exploring a Business Example Using Pandas 142
Pandas in Python 146
Import and View Data 147
Visualize the Data 149
Data Cleaning and Modification 153
Reading from Different Data Sources 155
Data Filtering and Grouping 161
Scikit-learn 167
Training and Predicting with a Linear Regression Model 168
Using a Random Forest Classifier 175
Training a Decision Tree 177
A Clustering Example (Unsupervised Learning) 182
Conclusion 183
6. TensorFlow and Keras for Predictive Analytics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
TensorFlow Fundamentals 186
Linear Regression Using TensorFlow 193
Data Preparation 194
Model Creation and Training 196
Predictions and Model Evaluation 200
Deep Neural Networks in TensorFlow 202
Conclusion 211
7. Predictive Analytics for Business Problem-Solving. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Prediction-Based Optimal Retail Price Recommendations 213
Using a Simple Linear Regression Model 216
Using a Polynomial Regression Model 221
Using Multivariate Regression 233
An Introduction to Recommender Systems 246
Building Recommender Systems Using surprise scikit in Python 250
Credit Card Fraud Classification 267
Credit Card Fraud Baseline Analysis Using Artificial Neural Networks 268
Credit Card Fraud Weighted Analysis Using Artificial Neural Networks 285
Credit Card Analysis with Multiple Hidden Layers
in the Artificial Neural Network 289
Conclusion 292
Table of Contents | vii
8. Exploring AWS Cloud Provider Services for AI/ML. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
To Cloud or Not to Cloud 293
Exploring AWS SageMaker 295
Prerequisites 296
Data Ingest and Exploration 297
Data Transformation 304
Model Training and Prediction 314
Cleanup 316
Exploring Amazon Forecast 317
Import Data 319
Train the Predictor 321
Create a Forecast 322
What-If Analysis 324
Cleanup 326
Conclusion 327
9. Food for Thought. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
A Few More Use Cases 331
Navigation and Traffic Management 331
Credit Scoring 333
The Social Impact of Predictions 334
Conclusion 336
Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
viii | Table of Contents