A Novel K-Means Hierarchical Clustering Algorithm for Efficient Information Extraction from Large Data Sets

摘要

Recent emphasis on exploratory data analysis targeted to identifying groups or structures in multivariate data has contributed to the development of a number of clustering techniques. Among these, hierarchical methods and K-means clustering have emerged as two popular approaches. In this paper, we shall present an algorithm that efficiently combines the strong points of these two approaches in a novel two-stage process that outperforms the individual components in both accuracy of representation and computation time. To make the combination more effective, we propose a new initialization scheme for the K-means stage to achieve improved codebook placement. We also propose a novel visualization scheme that combines the Principal Component Analysis (PCA) and Minimal Spanning Tree (MST) in an arrangement that ensures reliability of the visualization. The performance of the clustering and visualization techniques proposed here is illustrated by application to a challenging data set popularly used as a benchmark.