Improvement of Pearson similarity coefficient based on item frequency
Fie Zhang, Weitao Zhou, Lili Sun, Xiaochuan Lin, Hongjie Liu, Zhimin He · 2017
Personalized recommender system is increasingly used to overcome the problem of information overload, collaborative filtering recommender algorithm is one of the most popular algorithms, which provides recommendations by the information of the neighbors who have the same preferences of the target user. Because the purpose of recommender system is to provide personalized recommendations, so for a unpopular item which rated by little users, the recommend frequency should be reduced. While for the popular item which could not reflect the preferences of users, the recommend frequency should be decreased too. Therefore, base on the information of item frequency, an improved Pearson correlation method is proposed. Experimental results show that the proposed method can improve the quality of the recommendations.