Profile driven Collaborative Filtering

The problem of Recommender systems is to apply knowledge discovery techniques and to provide personalized recommendations. A recommender system tracks past actions of a group of users to make recommendations to individual members of the group. In today’s e-commerce world these systems are achieving widespread success. Content based and collaborative filtering based approaches are the two major approaches under which these systems are classified. The two are combined to form a hybrid approach. The major challenges before today’s recommender systems are to provide high quality recommendations and to provide these in a large-scale environment. In this paper we present a hybrid solution that remains scalable and efficient. Our system in a novel way makes use of content based and collaborative filtering approaches. We determine the profile of the user from the items he is interested in, and then apply collaborative filtering to determine his nearest neighbors with users of same profile. This way we exploit both content and collaborative filtering approaches. Frequent item-set construction algorithm was employed in determining the profile of the user and cosine similarity measure in determining the neighbors. We present the evaluation of the approach by recommending movies on Indian movie dataset.