An Empirical Evaluation of k-Means Clustering Technique and Comparison

Abstract

In this research work we have studied the behavior of k-means clustering technique and we have made an analysis. We have mentioned the limitations and some methods to improve it. Next we have discussed two improved versions of k-means algorithm - in the first version, a greedy method is being applied to overcome some of the limitation whereas in the second version, using some pre-computation, we can improve the traditional k-means technique to some extent. We have implemented these algorithms in python programming language and compared the results. We have also investigated the ability of four cluster validation technique in terms of these three clustering techniques. From the experimental outcomes, indications regarding the optimal validation method as well as optimal clustering algorithm is being presented.