Clustering of datasets using PSO-K-Means and PCA-K-means
Anusuya S. Venkatesan, Latha Parthiban · 2011
Cluster analysis plays indispensable role in obtain ing knowledge from data, being the first step in da ta mining and knowledge discovery. The purpose of data clustering is to rev eal the data patterns and gain some initial insight s regarding data distribution. K-means is one of the widely used partitional clust ering algorithms and it is more sensitive to outlie rs and do not work well with high dimensional data. In this paper, K-means has b een integrated with other approaches to overcome th e shortcomings hereby improving the accuracy of clustering. In this paper , basic k-means and the combination of k-means wit h PCA and PSO are applied on various datasets from UCI repository. Th e experimental results of this paper show that PSO- K-means and PCA-KMeans improves the performance of basic K-means in terms of accuracy and computational time .