Entropy-based Ranking Approach for Enhancing Diversity in Tag-based Community Recommendation

Abstract

Accuracy is the dominant performance evaluation in recommender systems. However, user satisfaction in recommendation includes other factors such diversity and novelty. Some solutions for improving diversity in recommendation, use re-ranking as a post-process on recommendation list achieved from accuracy-aware algorithms. In this work, we propose a method to involve entropy of communities as a diversity factor into the predicted weights from HOSVD method, helping to improve diversity in recommendation list without re-ranking. Experiments on Last.FM dataset, for the case of community recommendation with multi-mode data including users and tags for each community, proves the benefit on introducing diversity factor into the accuracy-based recommendation solutions to improve diversity.