Data Mining Algorithms: Explained Using R

Chapter 19

Attribute selection

19.1 Introduction

Unlike the classification, regression, or clustering tasks, the attribute selection task is not a data mining task per se, as—contrary to the former—it is not concerned with delivering models representing generalizations of predictively useful relationships discovered in the data. While the results of attribute selection—taking the form of a subset of attributes—can be interesting and insightful on their own—there is always another task (usually classification, regression, or clustering) for which attribute selection serves as preprocessing, and which provides the motivation, context, and quality criteria therefore. This puts attribute selection in the same category as data transformation techniques discussed in Chapter 17. What distinguishes attribute selection from the latter is that it often requires much more refined and computationally demanding algorithms to be adequately performed, and its impact on the final model quality may be even more substantial. It cannot be therefore reduced to something purely technical and trivial by any means, and definitely deserves more interest than it usually receives.

dmr.data

Example 19.1.1

Attribute selection algorithms presented in this chapter will be illustrated by R code examples, using the toy weather, weatherc, and weatherr datasets from Examples 1.3.1, 1.3.2, and 1.4.1, as well as the more realistic Vehicle Silhouettes, Soybean, and Boston Housing datasets from the ...