Automated machine learning for business

목차 (Table of Contents)

• CONTENTS
• Preface = xi
• Automated Machine Learning (AutoML) = xii
• A Note to Instructors = xii
• Acknowledgments = xiii
• Book Outline = xiii
• Dataset Download = xvii
• Copyrights = xvii
• SECTION Ⅰ WHY USE AUTOMATED MACHINE LEARNING?
• 1 What Is Machine Learning? = 3
• 1.1 Why Learn This? = 3
• 1.2 Machine Learning Is Everywhere = 4
• 1.3 What Is Machine Learning? = 6
• 1.4 Data for Machine Learning = 8
• 1.5 Exercises = 10
• 2 Automating Machine Learning = 11
• 2.1 What Is Automated Machine Learning? = 12
• 2.2 What Automated Machine Learning Is Not = 14
• 2.3 Available Tools and Platforms = 15
• 2.4 Eight Criteria for AutoML Excellence = 17
• 2.5 How Do the Fundamental Principles of Machine Learning and Artificial Intelligence Transfer to AutoML? A Point-by-Point Evaluation = 20
• 2.6 Exercises = 21
• SECTION Ⅱ DEFINING PROJECT OBJECTIVES
• 3 Specify Business Problem = 25
• 3.1 Why Start with a Business Problem? = 25
• 3.2 Problem Statements = 26
• 3.3 Exercises = 29
• 4 Acquire Subject Matter Expertise = 31
• 4.1 Importance of Subject Matter Expertise = 31
• 4.2 Exercises = 32
• 5 Define Prediction Target = 33
• 5.1 What Is a Prediction Target? = 33
• 5.2 How Is the Target Important for Machine Learning? = 35
• 5.3 Exercises / Discussion = 36
• 6 Decide on Unit of Analysis = 37
• 6.1 What Is a Unit of Analysis? = 37
• 6.2 How to Determine Unit of Analysis = 38
• 6.3 Exercises = 39
• 7 Success, Risk, and Continuation = 40
• 7.1 Identify Success Criteria = 40
• 7.2 Foresee Risks = 41
• 7.3 Decide Whether to Continue = 44
• 7.4 Exercises = 45
• SECTION Ⅲ ACQUIRE AND INTEGRATE DATA
• 8 Accessing and Storing Data = 51
• 8.1 Track Down Relevant Data = 51
• 8.2 Examine Data and Remove Columns = 54
• 8.3 Example Dataset = 55
• 8.4 Exercises = 58
• 9 Data Integration = 59
• 9.1 Joins = 60
• 9.2 Exercises = 69
• 10 Data Transformations = 70
• 10.1 Splitting and Extracting New Columns = 70
• 10.1.1 IF-THEN Statementsand One-hot Encoding = 70
• 10.1.2 Regular Expressions (RegEx) = 72
• 10.2 Transformations = 78
• 10.3 Exercises = 79
• 11 Summarization = 80
• 11.1 Summarize = 80
• 11.2 Crosstab = 84
• 11.3 Exercises = 87
• 12 Data Reduction and Splitting = 88
• 12.1 Unique Rows = 88
• 12.2 Filtering = 91
• 12.3 Combiningthedata = 92
• 12.4 Exercises = 94
• SECTION Ⅳ MODEL DATA
• 13 Startup Processes = 97
• 13.1 Uploading Data = 97
• 13.2 Exercise = 102
• 14 Feature Understanding and Selection = 103
• 14.1 Descriptive Statistics = 103
• 14.2 Data Types = 107
• 14.3 Evaluations of Feature Content = 110
• 14.4 MissingValues = 112
• 14.5 Exercises = 113
• 15 Build Candidate Models = 114
• 15.1 Startingthe Process = 114
• 15.2 Advanced Options = 116
• 15.3 Starting the Analytical Process = 121
• 15.4 Model Selection Process = 127
• 15.4.1 Tournament Round 1 : 32% Sample = 128
• 15.4.2 Tournament Round 2 : 64% Sample = 131
• 15.4.3 Tournament Round 3 : Cross Validation = 131
• 15.4.4 Tournament Round 4 : Blending = 132
• 15.5 Exercises = 133
• 16 Understanding the Process = 134
• 16.1 Learning Curves and Speed = 134
• 16.2 Accuracy Tradeoffs = 138
• 16.3 Blueprints = 139
• 16.3.1 Numeric Data Cleansing (Imputation) = 140
• 16.3.2 Standardization = 142
• 16.3.3 One-hot Encoding = 143
• 16.3.4 Ordinal Encoding = 147
• 16.3.5 Matrix of Word-gram Occurrences = 149
• 16.3.6 Classification = 151
• 16.4 Hyperparameter Optimization (Advanced Content) = 154
• 16.5 Exercises = 156
• 17 Evaluate Model Performance = 157
• 17.1 Introduction = 157
• 17.2 A Sample Algorithm and Model = 159
• 17.3 ROC Curve = 164
• 17.4 Usingthe Lift Chart and Profit Curve for Business Decisions = 176
• 17.5 Exercises = 179
• 18 Comparing Model Pairs = 180
• 18.1 Model Comparison = 180
• 18.2 Prioritizing Modeling Criteria and Selecting a Model = 185
• 18.3 Exercises = 187
• SECTION Ⅴ INTERPRET AND COMMUNICATE
• 19 Interpret Model = 191
• 19.1 Feature Impacts on Target = 191
• 19.2 The Overall Impact of Features on the Target without Consideration of Other Features = 192
• 19.3 The Overall Impact of a Feature Adjusted for the Impact of Other Features = 193
• 19.4 The Directional Impact of Features on Target = 194
• 19.5 The Partial Impact of Features on Target = 195
• 19.6 The Power of Language = 198
• 19.7 Hotspots = 201
• 19.8 Prediction Explanations = 203
• 19.9 Exercises = 205
• 20 Communicate Model Insights = 206
• 20.1 Unlocking Holdout = 207
• 20.2 Business Problem First = 209
• 20.3 Pre-processing and Model Quality Metrics = 210
• 20.4 Areas Where the Model Struggles = 213
• 20.5 Most Predictive Features = 214
• 20.6 Not All Features Are Created Equal = 214
• 20.7 Recommended Business Actions = 217
• 20.8 Exercises = 218
• SECTION Ⅵ IMPLEMENT, DOCUMENT AND MAINTAIN
• 21 Set Up Prediction System = 221
• 21.1 Retraining Model = 221
• 21.2 Choose Deployment Strategy = 222
• 21.3 Exercises = 227
• 22 Document Modeling Process for Reproducibility = 228
• 22.1 Model Documentation = 228
• 22.2 Exercises = 229
• 23 Create Model Monitoring and Maintenance Plan = 230
• 23.1 Potential Problems = 230
• 23.2 Strategies = 230
• 23.3 Exercises = 232
• 24 Seven Types of Target Leakage in Machine Learning and an Exercise = 233
• 24.1 Types of Target Leakage = 233
• 24.2 A Hands-on Exercise in Detecting Target Leakage = 236
• 24.3 Exercises = 239
• 25 Time-Aware Modeling = 240
• 25.1 An Example of Time-Aware Modeling = 240
• 25.1.1 Problem Statement = 240
• 25.1.2 Data = 241
• 25.1.3 Initialize Analysis = 241
• 25.1.4 Time-Aware Modeling Background = 241
• 25.1.5 Data Preparation = 244
• 25.1.6 Model Building and Residuals = 247
• 25.1.7 Candidate Models = 247
• 25.1.8 Selectingand Examininga Model = 249
• 25.1.9 A Small Detour into Residuals = 253
• 25.1.10 Model Value = 256
• 25.1.11 Learning about Avocado Price Drivers = 256
• 25.2 Exercises = 258
• 26 Time-Series Modeling = 259
• 26.1 The Assumptions of Time-Series Machine Learning = 259
• 26.2 A Hands-on Exercise in Time-Series Analysis = 260
• 26.2.1 Problem Context = 260
• 26.2.2 Loading Data = 262
• 26.2.3 Specify Time Unit and Generate Features = 262
• 26.2.3 Examine Candidate Models = 268
• 26.2.4 Digging into the Preferred Model = 270
• 26.2.5 Predicting = 273
• 26.3 Exercises = 275
• Appendix A. Datasets = 277
• A.1 Diabetes Patients Readmissions = 277
• Summary = 277
• Business Goal = 277
• Datasets = 277
• Exercises = 280
• Rights = 280
• A.2 Luxury Shoes = 280
• Summary = 280
• Business Goal = 281
• Datasets = 281
• Exercises = 283
• A.3 Boston Airbnb = 283
• Summary = 283
• Business Goal = 284
• Datasets = 284
• Rights = 287
• A.4 Part Backorders = 287
• Summary = 287
• Business Goal = 287
• Datasets = 287
• Exercises = 288
• Rights = 289
• A.5 Student Grades Portuguese = 289
• Summary = 289
• Business Goal = 289
• Datasets = 289
• Exercises = 290
• Rights = 293
• A.6 Lending Club = 293
• Summary = 293
• Business Goal = 294
• Dataset = 294
• Rights = 300
• A.7 College Starting Salaries = 300
• Summary = 300
• Business Goal = 300
• Datasets = 300
• Exercises = 301
• Rights = 301
• A.8 HR Attrition = 301
• Summary = 301
• Business Goal = 302
• Datasets = 302
• Exercises = 304
• Rights = 305
• A.9 Avocadopocalypse Now? = 305
• Summary = 305
• Business Goal = 306
• Datasets = 306
• Exercises = 307
• Rights = 307
• Appendix B. Optimization and Sorting Measures = 308
• Appendix C. More on Cross Validation = 311
• References = 315
• Index = 319