An Efficient Fuzzy Clustering for Web Log Mining

摘要

Similar access interest of Web session group can be discovered by analyzing the Log in the web site. The site designer can adjust the structure of the web site, or provide web personalization service according to the result of clustering. A new clustering algorithm of Web session is presented. We define the notion of a "user session" and a new distance measure between two web sessions. The objective is to minimize the overall inter-cluster dissimilarity, which is measured by the summation of fuzzy dissimilarity between the objects and the centroid of the cluster they are assigned to. The experiment shows that this approach is successful.