Collaborative filtering recommendation algorithm considering users’ preferences for item attributes

Abstract

In the neighborhood-based collaborative filtering recommendation algorithm, the accuracy of the similarity calculation determines the quality of the recommendation algorithm directly. The traditional similarity measure only considers influence of common rated items among users, and ignores the attribute characteristics of users’ rated items. Low-precision similarity metrics reduce performance of recommended systems, when the dataset is extremely sparse. In order to solve above problems, this paper proposes a similarity measure model considering users’ preferences for item attributes. The model fully considers the user’s preferences for item attributes and co-rated items, and the number of co-rated items. The model establishes more connections between users and items, so as to mine user interests effectively and make it more in line with the actual application. The experimental results show that the model proposed by this paper is superior to other comparison methods in accuracy and diversity, which effectively improves the performance of the recommended algorithm.