Multi-objective optimization for clustering microarray gene expression data - a comparative study

Abstract

Clustering is one of the main data mining tasks. It can be performed on a fuzzy or a crisp basis. Fuzzy clustering is widely-applied with microarray gene expression data as these data are usually uncertain and imprecise. There are several measures to evaluate the quality of clustering, but their performance is highly related to the dataset to which they are applied. In a previous work the authors proposed using a multi-objective genetic algorithm – based method, NSGA – II, to optimize two clustering validity measures simultaneously. In this paper we use another multi-objective optimizer, NSPSO, which is based on the particle swarm optimization algorithm, to solve the same problem. The experiments we conducted on two microarray gene expression data show that NSPSO is superior to NSGA-II in handling this problem.