Title:Hybrid algorithms to customize and optimize diversity and accuracy of recommendations

Abstract: Recommender systems use data on past user preferences to predict possible future likes and interests. A key challenge is how to obtain accurate recommendations while avoiding suggestions that are too obvious or insufficiently personalized. In this paper we take four recommendation algorithms and test them on three distinct datasets for accuracy and diversity of their results. Based on this we show how two of the methods can be combined into a hybrid that enables simultaneous gains in both accuracy and diversity of recommendations.