An improved web page recommendation system using partitioning and web usage mining

Abstract

There are different types of hypertext documents available on the Internet. Accessing relevant information and serving useful information to the user from the Internet has become a complex and expensive task. To make this process simpler, one of the widely used recommendation systems is item based collaborative filtering recommendation system which predicts web pages based on the browsing activity of the user on the Internet and recommends web pages as per their interests. There are certain challenges in these systems like sparsity and scalability, the proposed approach overcomes these problems. The proposed approach uses weighted kmean clustering instead of simple k-mean clustering and the obtained clusters are partitioned on the basis of similarity which helps in reducing the processing time of recommendation generation. Clustering and partitioning enhances the existing item based collaborative filtering recommendation system. The MovieLens data set is used for demonstrating the proposed approach. The performance of the proposed approach is evaluated using various metrics. The result shows that the proposed approach is 30% efficient in terms of root mean square error and 21% effective in respect of mean absolute error analysis and the accuracy measures factors like precision, recall and F-measure are found to have higher values than the existing item based collaborative filtering recommendation systems. � 2015 ACM.