Analysing exposure diversity in collaborative recommender systems—Entropy fusion approach

Abstract

Recommender Systems are considered as essential business tools to leverage the potential growth of on-line services. Neighbourhood based collaborative filtering, a successful recommendation approach has mainly focused on improving accuracy of predictions. From user point of view, it is more valuable to obtain novel and diverse recommendations rather than monotonic preferences. Ratings given by a user for different categories of items are considered as a tool to access user exposure diversity which signifies his creative and divergent thinking. On the other hand, pair of items is concordant if highly correlated users agree in rating the items. Based on the user exposure diversity and item concordance, the neighbourhood selection process of item based collaborative recommender systems is refined. Rating predictions are made based on the newly selected neighbours. The performance of the proposed approach is investigated for accuracy and diversity of predictions on Movielens data sets. The results demonstrate that the proposed approach outperforms the state of the art recommendation approaches which address accuracy–diversity trade off. Statistical analysis is done to prove the efficiency of the proposed approach.