A collaborative filtering algorithm based on user interest and joint similarity

Abstract：In recommender system, sparsity and timeliness are the usually faced problems.A new multiplied similarity computation with time-information was proposed, which applied to user-based collaborative filtering method, to better reflect the neighbor dependency of different users at different stages, then to dig out the personalized information that implied in the rating items.Firstly, the fusion of joint similarity method, collaborative similarity, preference similarity and trajectory similarity were addressed to gain the users neighbor numbers by adjust the relative weight and the threshold of similarity.Then, the time-weight Logistic function in the calculation process of joint similarity was used to improve the timeliness.Finally, the experiments were carried out.The results showed that the proposed method had better recommendation accuracy than the classical algorithms, and could more effectively predict the user's true rating.

Keyword：joint similarity;sparsity;time-weight function;collaborative filtering algorithm;