Improving diversity performance of association rule based recommender systems

Abstract

In recommender systems (RSs), getting high accuracy alone does not improve the user satisfaction. In the literature, efforts are being made to improve the variety/diversity of recommendations for higher user satisfaction. In these efforts, the accuracy performance is reduced at the cost of improving the variety of recommendations. In this paper, we propose an approach to improve the diversity as well as accuracy of association rule based RSs. We propose a ranking mechanism, called diverse rank, to rank association rules based on the diversity of the items in the pattern. The recommendations are made based on association rules with high confidence and diversity. The experimental results on the realworld MovieLens data set show that the proposed approach improves the performance of association rule based RSs with better diversity without compromising the accuracy.