A two-stage genetic k-harmonic means method for data clustering

Abstract

Clustering techniques are aimed to partition the entire input space into disconnected sets where the members of each set are highly connected. Kharmonic means (KHM) is a well-known data clustering technique, but it runs into local optima. A two stage genetic clustering method using KHM (TSGKHM) is proposed in this research, which can automatically cluster the input data points into an appropriate number of clusters. With the best features of both the algorithm, and TSGKHM the first stage overcomes the local optima and results in optimal cluster centers, and in the second stage, results into/in optimal clusters. The proposed method is executed on globally accepted, four real time data sets. The intermediate results are produced. The performance analysis shows that TSGKHM performs significantly better.