Improving collaborative recommender systems via emotional features
Soghra Lazemi, Hossein Ebrahimpour-Komleh · 2016
Nowadays, by communication networks expansion, recommender systems play an important role in our daily life. Recommender system tries to recommend items that are attractive and pleasant for user. At the same time, collaborative filtering approaches is one of the most successful approaches which recommend to the customer other users pleasant items which have similar interests with him/her in the past. In this paper, we present a novel approach of these systems by entering user's emotions in recommender system. Then, by emotion matrix definition, the users who are similar emotionally are searched to improve the performance of traditional user-based collaborative filtering (UBCF). The proposed algorithm offers items to him/her by considering user's current emotion which users with similar emotion like them. This algorithm identifies suitable number of neighbors automatically and chooses high-quality neighbors to do predictions by applying pre-processing phase. Our proposed algorithm has a solution for user cold-start problem. Evolution results in Jester database show that proposed method has better prediction accuracy and makes high quality recommendations.