A GP based approach to the classification of multiclass microarray datasets

Abstract

In this paper, we propose a genetic programming (GP) based approach to analyze multiclass microarray datasets. Here, a multiclass problem is divided into a set of two-class problems. Instead of applying a tree for each two-class problem, a small-scale ensemble system containing a set of trees is deployed and denoted by sub-ensemble (SE). The SEs tackling the respective two-class problems are combined to construct an individual of the GP, so that an individual can deal with a multiclass problem directly. In the experiments, the GP implements classification and feature selection at the same time. The results obtained at independent test sets show that our method is efficient in the search of genes with great biological significance, and achieves high classification accuracy at the same time. © 2008 Springer-Verlag Berlin Heidelberg.