Categorical feature reduction using multi objective genetic Algorithm in cluster analysis

Abstract

In the paper, real coded multi objective genetic algorithm based K-clustering method has been studied, K represents the number of clusters. In K-clustering algorithm value of K is known. The searching power of Genetic Algorithm (GA) is exploited to search for suitable clusters and centers of clusters so that intra-cluster distance (Homogeneity, H) and inter-cluster distances (Separation, S) are simultaneously optimized. It is achieved by measuring H and S using Mod distance per feature metric, suitable for categorical features (attributes). We have selected 3 benchmark data sets from UCI Machine Learning Repository containing categorical features only. The paper proposes two versions of MOGA based K-clustering algorithm. In proposed MOGA (H, S), all features are taking part in building chromosomes and calculation of H and S values. In MOGA-Feature-Selection (H, S), selected features take part to build chromosomes, relevant for clusters. Here, K-modes is hybridized with GA. We have used hybridized GA to combine global searching capabilities of GA with local searching capabilities of K-modes. Considering context sensitivity, we have used a special crossover operator called "pairwise crossover" and "substitution". The main contribution of this paper is simultaneous dimensionality reduction and optimization of objectives using MOGA. © 2013 Springer-Verlag Berlin Heidelberg.