Comparative Analysis of Web Usage Data Clustering Using Asymmetric Binary Variables and K-Means

Abstract

World Wide Web overwhelms us with the immense amounts of widely distributed interconnected, rich and dynamic information. As a consequence of this, Web Usage Mining becomes one of the popular research areas. It involves the application of data mining techniques to discover usage patterns from the Web access logs data. Clustering is one of the important functions in Web Usage Mining to group the user access patterns which have the same access behavior. In this paper, we would like to propose a new approach, asymmetric binary variables (one type of Jaccard coefficient) to perform clustering. And then the performance of our proposed approach is compared with k-means clustering. The resulting clusters from these two methods are tested with two internal validation methods: Dunn Index and DB Index (Davies and Bouldin Index). Finally, we point out the strengths and weaknesses of each method. According to the analysis results, the findings of clustering upon these methods can be seen clearly.