Supervised machine learning: optimization framework and applications with SAS and R

Cover......Page 1
Half Title......Page 2
Title Page......Page 4
Copyright Page......Page 5
Dedication......Page 6
Table of Contents......Page 8
Acknowledgments......Page 14
Authors......Page 16
Introduction: Challenges in the Application of Machine Learning Classification Methods......Page 18
Part I......Page 24
  1.1 Components of the AI Framework and Their Interaction......Page 26
   1.2.1 Creating Training and Test Datasets......Page 28
   1.2.4 Data Contamination......Page 29
  1.4 R Libraries for the AI Framework Components......Page 30
  References......Page 31
  2.1 Introduction......Page 32
  2.2 Principles of Supervised Machine Learning......Page 33
   2.3.1 Introduction......Page 35
    2.3.2.1 Activation Function......Page 36
    2.3.2.2 Neurons......Page 37
    2.3.2.3 Networks......Page 38
   2.4.1 Introduction......Page 39
   2.4.2 Kernel......Page 41
   2.4.3 Margin......Page 42
   2.4.5 Bias–Variance Trade-off and SVM Hyperparameters......Page 43
   2.4.6 R for SVM......Page 44
   2.5.1 Introduction......Page 45
   2.5.3 Problem of Separation......Page 46
   2.5.5 SAS for Firth’s Regression......Page 48
  References......Page 50
  3.1 Introduction......Page 52
  3.2 Overview of Bootstrap Methods......Page 53
   3.2.1 The Basic Bootstrap......Page 54
   3.2.2 Hypothesis Tests, Estimates, and Confidence Intervals......Page 55
   3.2.3 Bias Reduction......Page 57
   3.2.4 The Parametric Bootstrap......Page 58
   3.2.6 Bootstrap Samples Similarity......Page 59
   3.3.1 m-out-of-n in SAS......Page 60
   3.3.2 m-out-of-n in R......Page 61
  3.4 Summary......Page 62
  References......Page 63
  4.1 Introduction......Page 64
   4.2.1 Minimum Covariance Determinant Method......Page 65
   4.2.2 MCD in SAS......Page 66
  4.3 Bias Reduction......Page 67
  References......Page 68
  5.1 Introduction......Page 70
  5.2 Application of DoE in AI Framework......Page 71
    5.2.2.1 Randomization......Page 72
   5.2.3 Full-Factorial Experiment......Page 73
   5.2.4 Fractional Factorial Experiment......Page 80
   5.2.5 Linear Mixed Models......Page 81
   5.2.6 Factors and Response Variables in the AI Framework......Page 82
   5.2.7 Example......Page 83
   5.2.8 Analysis of Linear Mixed Model Using SAS......Page 85
   5.2.9 Analysis of Linear Mixed Model Using R......Page 88
  5.3 Summary......Page 89
  References......Page 90
Part II......Page 92
  6.1 Principles of Code-Free Design......Page 94
   6.2.1 Relational Model......Page 95
   6.2.4 Relational Data Structure......Page 96
   6.2.7 Functions......Page 97
   6.2.12 Missing Values......Page 98
   6.3.1 The Library Table......Page 99
   6.3.4 The Message Table......Page 100
   6.3.5 The Property Table......Page 101
   6.3.8 Process of Application Data Model Definition......Page 102
   6.3.9 Features of the Data Dictionary......Page 103
   6.4.1 How to Generate Application Objects......Page 104
   6.4.2 Generating R Datasets from the Data Dictionary Metadata......Page 107
   6.4.3 SAS and R Interoperability......Page 109
  Reference......Page 111
    7.1.1.1 The Input Data Dictionary......Page 112
    7.1.1.2 Input and Structure Tables......Page 113
    7.1.1.3 Outlier_Detection and Bias_Correction Tables......Page 114
    7.1.1.4 Bootstrap Table......Page 115
    7.1.1.5 Output Table......Page 116
   7.1.2 SAS Macro Program......Page 118
  7.2 Summary......Page 121
  8.1 Data Dictionary......Page 122
   8.1.2 Features Table......Page 123
   8.1.3 Metrics Table......Page 124
   8.1.6 Results Table......Page 125
   8.1.7 Results_Metrics Table......Page 126
  8.3 R Programs......Page 128
  Reference......Page 129
  9.1 Data Dictionary......Page 130
   9.1.1 Contamination Table......Page 131
   9.1.3 Cont_Results Table......Page 132
  9.3 R Programs......Page 133
  Reference......Page 134
Part III......Page 136
  10.1 Introduction......Page 138
    10.2.1.2 Conversion Rate......Page 139
    10.2.1.3 Dynamic Conversion Rate......Page 140
    10.2.1.4 Time-to-Deal......Page 141
    10.2.2.2 Application Flow......Page 142
    10.2.2.3 Dynamic Conversion Rate per Underwriter......Page 144
    10.2.2.4 Time-to-Deal per Underwriter......Page 145
  10.3 Traditional Approach to Knowledge Delivery......Page 146
   10.4.1 Data Structure......Page 147
   10.4.3 Bias–Variance Trade-Off and SVM Hyperparameters......Page 148
   10.4.4 Building the Classifier......Page 150
   10.4.6 Experimental Results......Page 153
  References......Page 155
  11.1 Introduction......Page 158
  11.3 The Cox Model for Claims Event Analysis......Page 159
  11.4 Application of the Cox Model for Claims Analysis......Page 161
   11.4.1 Data Transformation......Page 162
   11.4.2 Cox Model Assumption Validation......Page 164
   11.4.4 Deployment with SAS......Page 167
   11.4.5 Interpretation of Results......Page 169
  11.5 Summary......Page 175
  References......Page 176
Index......Page 178