Data Mining Algorithms: Explained Using R

Chapter 9

Regression trees

9.1 Introduction

Similarly as regression is sometimes referred to as “classification with continuous classes,” regression trees can be considered decision trees with continuous classes. Actually, the latter is much more justified than the former, since the differences between regression trees and decision trees are pretty minor compared to the differences between regression algorithms and classification algorithms in general. They can be thought of as slightly different instantiations of the same family of modeling algorithms based on hierarchical domain decomposition. They are discussed separately in this chapter rather than in Chapter 3 to preserve the task-oriented organization of this book and to better expose their specificity. To highlight the differences between regression trees and decision trees, the latter will be referred to for whatever they have in common, and this chapter will mostly focus on model representation and algorithm modifications that are necessary for the regression task.

Ex. 1.4.1dmr.data

Example 9.1.1

A number of examples will be presented throughout this chapter illustrating the major operations performed during regression tree with simple R code snippets. They closely follow the pattern of decision tree examples from Chapter 3. The weatherr data will be used for these examples. Besides auxiliary functions from DMR packages, the rpart package providing the standard R regression tree implementation will be used. ...