Weighted-Frequent Itemset Refinement Methodology (W-FIRM) of usage clusters

Abstract

Due to information overload on the Internet a large number of systems have been developed for extracting user behavior. This paper presents mining of Frequent Itemsets and refinement of usage clusters for web based applications. Here a particular case is under consideration where sessions in a cluster are in abundance, consequently leading to a very large number of not-so interesting recommendations for the user. To solve such problems we intend to refine clusters on the basis of Weighted Frequent Itemsets that in turn help to generate improved quality refined clusters. In the proposed work, Frequent Itemsets are sets of web pages that occur in sessions more than a given threshold known as the minimum support. Motivation for adapting Frequent Itemsets for refinement is the demand of dimensionality reduction. Experimental results show that the cluster quality using the proposed approach is better than the existing approaches (DBS, 2011 and HITS, 2010). After getting refined clusters the same can be used for number of applications such as Web Personalization, improvement in Web Site Structure, Analysis of Users' Online Behavior and the services of a Recommender System. © 2014 Springer International Publishing.