A comprehensive approach towards user-based collaborative filtering recommender system

Abstract

Recommender system refers to an information system that predicts the intuition of user observing behavior of all the users. The idea of collaborative filtering lies in producing a set of recommendations based on similarity as well as knowledge of users' relationships to items. In this paper, we combine some traditional similarity metrics to find three types of similar users which are super similar, super dissimilar and average similar. We also introduce a new similarity metric which is used in case of average similar user pairs effectively. Finally we evaluate the proposed method for recommendation by experimenting with real data of Movielens as well as Epinions. Thus we can conclude that our proposed similarity metric paves the way to take a comprehensive approach towards user-based collaborative filtering recommender system and performs better than other traditional similarity metrics.