Just Enough R!

An Interactive Approach to Machine Learning and Analytics

Richard J. Roiger

• Editore: Chapman and Hall/CRC
• Anno: 2020
• ISBN:
  9781000073560
• Edizione: 1
• Argomento: Computer Science
• Categoria: Data Processing

Just Enough R! An Interactive Approach to Machine Learning and Analytics presents just enough of the R language, machine learning algorithms, statistical methodology, and analytics for the reader to learn how to find interesting structure in data. The approach might be called "seeing then doing" as it first gives step-by-step explanations using simple, understandable examples of how the various machine learning algorithms work independent of any programming language. This is followed by detailed scripts written in R that apply the algorithms to solve nontrivial problems with real data. The script code is provided, allowing the reader to execute the scripts as they study the explanations given in the text.

Features

• Gets you quickly using R as a problem-solving tool

• 
  
  

• Uses RStudio's integrated development environment

• 
  
  

• Shows how to interface R with SQLite

• 
  
  

• Includes examples using R's Rattle graphical user interface

• 
  
  

• Requires no prior knowledge of R, machine learning, or computer programming

• 
  
  

• Offers over 50 scripts written in R, including several problem-solving templates that, with slight modification, can be used again and again

• 
  
  

• Covers the most popular machine learning techniques, including ensemble-based methods and logistic regression

• 
  
  

• Includes end-of-chapter exercises, many of which can be solved by modifying existing scripts

• 
  
  

• Includes datasets from several areas, including business, health and medicine, and science

• 
  
  

About the Author

Richard J. Roiger is a professor emeritus at Minnesota State University, Mankato, where he taught and performed research in the Computer and Information Science Department for over 30 years.

CHAPTER 1

Introduction to Machine Learning

In This Chapter

• Definitions and Terminology
• Machine Learning Strategies
• Evaluation Techniques
• Ethical Issues

THE R LANGUAGE CONTINUES to maintain its status as one of the top-rated problem-solving tools within the areas of machine learning, data science, data analytics, data mining, and statistical analysis. It’s easy to see why: R is free, contains thousands of packages, is supported by a growing community of users, and is easy to use when interfaced with RStudio’s integrated development environment!

R’s popularity has resulted in the development of thousands of tutorials on machine learning. The information is all there! Unfortunately, it’s easy to get lost in a maze of too much information. Valuable time is spent trying to find exactly what is needed to solve problems. The end result is frustration and difficulty understanding what’s important.

We believe our approach of presenting and clearly explaining script-based problem-solving techniques provides the tools you need for machine learning with R. The book’s title reflects its purpose. Just Enough R! gives you just enough of the R language and machine learning methods to minimize stumbling blocks and cut through the maze. Our goal is to give you what you need to become productive with R as quickly as possible.

In this chapter, we offer a brief introduction to machine learning. In Chapter 2, we move right into the nuts and bolts of the R language and the problem-solving techniques it offers. We conclude this chapter with a short summary, key term definitions, and a set of exercises. Let’s get started!

1.1 Machine Learning, Statistical Analysis, and Data Science

It’s almost impossible to surf the Web, open a newspaper, or turn on the TV without being exposed to terms such as machine learning, statistical analysis, data science, data analytics, and data mining. Most people have some idea about what these terms mean, but if you ask for a precise definition of any of them, you get a variety of answers. Here are a few distinctions:

• Building models to find structure in data has its roots in the fields of mathematics and statistics. Statistical methods are differentiated from other techniques in that they make certain assumptions about the nature of the data. Technically, if these assumptions are violated, the models built with these techniques may be inaccurate.
• Machine learning can be differentiated from statistical modeling in that assumptions about data distributions and variable independence are not a concern. Machine learning is generally considered an area of specialization within the broader field of artificial intelligence. However, most textbooks make little or no distinction between machine learning and statistical methods.
• Data science or data analytics is often defined as the process of extracting meaningful knowledge from data. Its methods come from several disciplines including computer science, mathematics, statistics, data warehousing, and distributed processing to name a few. Although machine learning is often seen in data science applications, it is not required.
• Data mining first became popular in the academic community about 1995 and can be defined as the process of using one or several machine learning algorithms to find structure in data. The structure may take many forms including a set of rules, a graph or network, a tree, one or several equations, and more. The structure can be part of a complex visual dashboard or as simple as a list of political candidates and an associated number representing voter sentiment based on twitter feeds.
• The phrase knowledge discovery in databases (KDD) was coined in 1989 to emphasize that knowledge can be derived from data-driven discovery and is frequently used interchangeably with data mining. In addition to performing data mining, a typical KDD process model includes a methodology for extracting and preparing data as well as making decisions about actions to be taken once data mining has taken place. As much of today’s data is not found in a traditional data warehouse, KDD is most often associated with knowledge discovery in data.

Although these general distinctions might be made, the most important point is that all of these terms define techniques designed to solve problems by finding interesting structure in data. We prefer to use the term machine learning as our focus is both on how to apply the algorithms and on understanding how the algorithms work. However, we often interchange the terms machine learning and data mining.

1.2 Machine Learning: A First Example

Supervised learning is probably the best and most widely used technique for machine learning. The purpose of supervised learning is twofold. First, we use supervised learning to build classification models from sets of data containing examples and nonexamples of the concepts to be learned. Each example or nonexample is known as an instance of data. Second, once a classification model has been constructed, the model is used to determine the classificat...