Clustering Based on PSO-K-Means and QPSO

Cluster analysis is typically the first step in data mining and knowledge discovery. The purpose of data clustering is to reveal the data patterns and gain some initial insights regarding data distribution. This paper focuses on exploring the applicability of PSO-K-Means and QPSO on Iris, Wine, Breast Cancer and Yeast datasets from UCI repository. PSO-K-Means is robust to outliers, benefits from both local and global view on data and overcomes the drawbacks of basic K-Means. The experimental results of this paper show that PSO-K-Means improves the performance of basic K-Means in terms of accuracy and computational time. The inter cluster and intra cluster distances using QPSO are found to be better compared to other standards like PSO and guarantees global convergence. The effectiveness of QPSO and PSO-K-Means are observed over PSO and K-Means for three different datasets and encouraging results are obtained.