Supervised Machine Learning

Optimization Framework and Applications with SAS and R

Tanya Kolosova, Samuel Berestizhevsky

• Publisher: Chapman and Hall/CRC
• Year: 2020
• eBook ISBN: 9781000176834
• Edition: 1
• Topic: Computer Science
• Subtopic: Statistics for Business & Economics
• Index: Computer Science

AI framework intended to solve a problem of bias-variance tradeoff for supervised learning methods in real-life applications. The AI framework comprises of bootstrapping to create multiple training and testing data sets with various characteristics, design and analysis of statistical experiments to identify optimal feature subsets and optimal hyper-parameters for ML methods, data contamination to test for the robustness of the classifiers.

Key Features:

• 
  
  

• Using ML methods by itself doesn't ensure building classifiers that generalize well for new data

• 
  
  

• Identifying optimal feature subsets and hyper-parameters of ML methods can be resolved using design and analysis of statistical experiments

• 
  
  

• Using a bootstrapping approach to massive sampling of training and tests datasets with various data characteristics (e.g.: contaminated training sets) allows dealing with bias

• 
  
  

• Developing of SAS-based table-driven environment allows managing all meta-data related to the proposed AI framework and creating interoperability with R libraries to accomplish variety of statistical and machine-learning tasks

• 
  
  

• Computer programs in R and SAS that create AI framework are available on GitHub

Part I

1

Introduction to the AI Framework

The artificial intelligence (AI) framework should be algorithm and technology agnostic. This means that it should not focus on a specific machine learning methodology and specific systems or software, but it applies to the design, application, and use of AI in general.

In this book, we describe the AI framework that helps to create machine learning classifiers in such a way that they produce a sufficiently good classification of unseen data. The framework incorporates data dictionaries that define processes such as statistically designed experiments, bootstrapping, and data “contamination.” This framework is fully transparent as all data processes are defined as metadata that ensures comparability and reproducibility of the results of the supervised classification.

Because the AI framework is fully independent of technology, it can be implemented using different software platforms. In this book, we present the implementation of the AI framework using two different software tools: SAS and R.

SAS is an integrated software system that enables accessing data across multiple sources, manipulating data, performing sophisticated analyses, and much more, and is one of the major choices of corporate business, though a bit costly.

R (R Core Team, 2018) is an open-source programming language developed by researchers, and many latest statistical and analytical methods are developed with R. This language has extensive documentation along with a supportive community and is widely used among statisticians and data scientists for developing statistical software and data analysis.

1.1 Components of the AI Framework and Their Interaction

The main idea of the AI framework proposed in this book is to create a systematic approach that resolves the challenges of machine learning modeling.

The process of building a machine learning classifier requires making choices that very often cannot be informed by intuition, understanding of a business problem, or pre-analysis of data. For example, making a decision about the feature subset requires considering different hypotheses, and it is reasonable to consider several different feature sets. Which one of them to choose eventually requires to go through the modeling process and comparison of classifiers built on different feature sets.

Choosing a machine learning method is challenging as well, and in many instances, the decision is made based on the convenience of the method and expected interpretability of the results, rather than based on theoretical and technical aspects of machine learning methods. If we do not want to limit ourselves to only one machine learning method, but rather want to tr...

Table of 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
• Part II
• Part III
• Index