An effective Web page recommender using binary data clustering

Abstract

Through growth of the Web, the amount of data on the net is growing in an uncontrolled way, that makes it hard for the users to find the relevant and required information- an issue which is usually referred to as information overload. Recommender systems are among the appealing methods that can handle this problem effectively. Theses systems are either based on collaborative filtering and content based approaches, or rely on rating of items and the behavior of the users to generate customized recommendations. In this paper we propose an efficient Web page recommender by exploiting session data of users. To this end, we propose a novel clustering algorithm to partition the binary session data into a fixed number of clusters and utilize the partitioned sessions to make recommendations. The proposed binary clustering algorithm is scalable and employs a novel method to find the representative of a set of binary vectors to represent the center of clusters--that might be interesting in its own right. In addition, the proposed clustering algorithm is integrated with the $$k$$k-means algorithm to achieve better clustering quality by combining its explorative power with fine-tuning power of the $$k$$k-means algorithm. We have performed extensive experiments on a real dataset to demonstrate the advantages of proposed binary data clustering methods and Web page recommendation algorithm. In particular, the proposed recommender system is compared to previously published works in terms of minimum frequency and based on the number of recommended pages to show its superiority in terms of accuracy, coverage and F-measure.