Genetic Programming for Classification and Algorithm Design

Abstract

Chapter 4 (entitled “Genetic Programming for Classification and Algorithm Design”), consists of two broad parts. The first part is about the classification task of data mining. In this first part, the chapter first explains important differences between classification models and classification algorithms - a crucial point to understand the contribution of this book, since the proposed genetic programming system produces a classification algorithm, rather than a classification model as usual in the literature. Then this chapter reviews genetic programming algorithms for evolving classification models, for evolving components of rule induction algorithms and for evolving classification systems as a whole (as a combination of datasets plus a classification algorithm). The second part of the chapter is about genetic programming for evolving the design of combinatorial optimization algorithms, rather than classification algorithms. Although combinatorial optimization is not the focus of this book, this topic was included in this chapter because the research on automatically evolving combinatorial optimization algorithms seems to be in a more advanced stage than the research on automatically evolving data mining algorithms, so that lessons from the former can be useful to the latter.