Supervised machine learning : optimization framework and applications with SAS and R / Tanya Kolosova and Samuel Berestizhevsky.

Format:

Book

Author/Creator:

Kolosova, Tanya, author.

Berestizhevsky, Samuel, author.

Language:

English

Subjects (All):

Supervised learning (Machine learning).

Physical Description:

1 online resource (xxiv, 160 pages)

Edition:

1st ed.

Place of Publication:

Boca Raton, Florida ; London ; New York : CRC Press, [2021]

Summary:

AI framework intended to solve a problem of bias-variance tradeoff for supervised learning methods in real-life applications. It comprises of bootstrapping to create multiple training and testing data sets, design and analysis of statistical experiments and optimal hyper-parameters for ML methods.

Contents:

Cover

Half Title

Title Page

Copyright Page

Dedication

Table of Contents

Acknowledgments

Authors

Introduction: Challenges in the Application of Machine Learning Classification Methods

Part I

1. Introduction to the AI Framework

1.1 Components of the AI Framework and Their Interaction

1.2 AI Framework in Detail

1.2.1 Creating Training and Test Datasets

1.2.2 Design of Experiments for a Classifier

1.2.3 Firth Logistic Regression

1.2.4 Data Contamination

1.2.5 Best Classifiers

1.3 SAS Procedures for the AI Framework Components

1.4 R Libraries for the AI Framework Components

References

2. Supervised Machine Learning and Its Deployment in SAS and R

2.1 Introduction

2.2 Principles of Supervised Machine Learning

2.3 Neural Network

2.3.1 Introduction

2.3.2 Neural Network Components

2.3.2.1 Activation Function

2.3.2.2 Neurons

2.3.2.3 Networks

2.3.3 R for Neural Networks

2.4 Support Vector Machine

2.4.1 Introduction

2.4.2 Kernel

2.4.3 Margin

2.4.4 Optimization

2.4.5 Bias-Variance Trade-off and SVM Hyperparameters

2.4.6 R for SVM

2.5 SVM Modification Using Firth's Regression

2.5.1 Introduction

2.5.2 Logistic Regression

2.5.3 Problem of Separation

2.5.4 R for Firth's Regression

2.5.5 SAS for Firth's Regression

2.6 Summary

3. Bootstrap Methods and Their Deployment in SAS and R

3.1 Introduction

3.2 Overview of Bootstrap Methods

3.2.1 The Basic Bootstrap

3.2.2 Hypothesis Tests, Estimates, and Confidence Intervals

3.2.3 Bias Reduction

3.2.4 The Parametric Bootstrap

3.2.5 m-out-of-n Bootstrap

3.2.6 Bootstrap Samples Similarity

3.3 Implementation of Bootstrap in SAS and R

3.3.1 m-out-of-n in SAS

3.3.2 m-out-of-n in R

3.4 Summary

References.

4. Outliers Detection and Its Deployment in SAS and R

4.1 Introduction

4.2 Outliers Detection and Treatment

4.2.1 Minimum Covariance Determinant Method

4.2.2 MCD in SAS

4.3 Bias Reduction

4.4 Summary

5. Design of Experiments and Its Deployment in SAS and R

5.1 Introduction

5.2 Application of DoE in AI Framework

5.2.1 Terminology of DoE

5.2.1.1 Experiment

5.2.1.2 Experimental Unit

5.2.1.3 Factor

5.2.1.4 Treatment

5.2.2 Principles of DoE

5.2.2.1 Randomization

5.2.2.2 Statistical Replication

5.2.2.3 Blocking

5.2.2.4 Orthogonality

5.2.3 Full-Factorial Experiment

5.2.4 Fractional Factorial Experiment

5.2.5 Linear Mixed Models

5.2.6 Factors and Response Variables in the AI Framework

5.2.7 Example

5.2.8 Analysis of Linear Mixed Model Using SAS

5.2.9 Analysis of Linear Mixed Model Using R

5.3 Summary

Part II

6. Introduction to the SAS- and R-Based Table-Driven Environment

6.1 Principles of Code-Free Design

6.2 The Data Dictionary Components for the AI Framework

6.2.1 Relational Model

6.2.2 Table

6.2.3 Data Aspects

6.2.4 Relational Data Structure

6.2.5 Domains

6.2.6 Relations and Tables

6.2.7 Functions

6.2.8 One-to-one Relationship

6.2.9 One-to-many Relationship

6.2.10 Primary Key

6.2.11 Foreign Key

6.2.12 Missing Values

6.2.13 Data Dictionary

6.3 Properties of the Data Dictionary

6.3.1 The Library Table

6.3.2 The Object Table

6.3.3 The Location Table

6.3.4 The Message Table

6.3.5 The Property Table

6.3.6 Meaning

6.3.7 The Link Table

6.3.8 Process of Application Data Model Definition

6.3.9 Features of the Data Dictionary

6.3.10 The Components of the Optimization Framework and Their Definitions in the Data Dictionary

6.4 Deployment of Code-Free Design with SAS and R.

6.4.1 How to Generate Application Objects

6.4.2 Generating R Datasets from the Data Dictionary Metadata

6.4.3 SAS and R Interoperability

6.5 Summary

Reference

7. Input Data Component

7.1 Overview of Data Management

7.1.1 Data Dictionary

7.1.1.1 The Input Data Dictionary

7.1.1.2 Input and Structure Tables

7.1.1.3 Outlier_Detection and Bias_Correction Tables

7.1.1.4 Bootstrap Table

7.1.1.5 Output Table

7.1.2 SAS Macro Program

7.1.3 R Program

7.2 Summary

8. Design of Experiment for Machine Learning Component

8.1 Data Dictionary

8.1.1 Experiment Table

8.1.2 Features Table

8.1.3 Metrics Table

8.1.4 ML_Method Table

8.1.5 Hyperparameters_Domain Table

8.1.6 Results Table

8.1.7 Results_Metrics Table

8.2 SAS Macro Program

8.3 R Programs

8.4 Summary

9. "Contaminated" Training Datasets Component

9.1 Data Dictionary

9.1.1 Contamination Table

9.1.2 Cont_Experiment Table

9.1.3 Cont_Results Table

9.1.4 Cont_Metric Table

9.2 SAS Macro Program

9.3 R Programs

9.4 Summary

Part III

10. Insurance Industry: Underwriters' Decision-Making Process

10.1 Introduction

10.2 Review of Underwriters' Performance

10.2.1 Metrics of Underwriters' Performance

10.2.1.1 Hit Ratio

10.2.1.2 Conversion Rate

10.2.1.3 Dynamic Conversion Rate

10.2.1.4 Time-to-Deal

10.2.2 Analysis of Underwriters' Performance

10.2.2.1 Data Description

10.2.2.2 Application Flow

10.2.2.3 Dynamic Conversion Rate per Underwriter

10.2.2.4 Time-to-Deal per Underwriter

10.3 Traditional Approach to Knowledge Delivery

10.4 Anatomy of Artificial Intelligence Solution

10.4.1 Data Structure

10.4.2 Classification Approach

10.4.3 Bias-Variance Trade-Off and SVM Hyperparameters

10.4.4 Building the Classifier.

10.4.5 "Contamination" of Training Datasets

10.4.6 Experimental Results

10.5 Summary

11. Insurance Industry: Claims Modeling and Prediction

11.1 Introduction

11.2 Data

11.3 The Cox Model for Claims Event Analysis

11.4 Application of the Cox Model for Claims Analysis

11.4.1 Data Transformation

11.4.2 Cox Model Assumption Validation

11.4.3 Bayesian Machine Learning Approach

11.4.4 Deployment with SAS

11.4.5 Interpretation of Results

11.5 Summary

Index.

Notes:

"A Chapman & Hall book"--Title page.

Description based on print version record.

ISBN:

0-429-29759-9

1-000-17681-9