Grouping of users based on user navigation behaviour using supervised association rule tree mining

Abstract

In this internet world, an increased interest of users in search of World Wide Web results in wide relevance of web mining, an application of data mining. Clustering has been widely used for web usage mining. Finding initial cluster center and specifying the number of clusters are the major challenges, which are overcome in this work by grouping of users based on the target class value. The benchmark dataset MSNBC is collected for the entire day of September 28, 1999. Supervised association rule tree mining is used to find frequent itemset for the targeted class value and thus generating 'if then rules'. Users are automatically clustered based on the rules satisfying the ground truth, resulting in 36 clusters in two iterations. The results revealed that the renowned clustering algorithms such as K-means takes 22 iterations for forming 36 clusters, wherein the proposed work generates 36 clusters in two iterations.