Personalized Recommendation Algorithm Considering Time Sensitivity

Aiming to solve the problem of goods popularity bias, this paper introduces the prevalence of items into user interest modeling, and proposes an item popularity model based on user interest feature. Usually, traditional model that does not take into account the stability of user’s interests, which leads to the difficulty in capturing their interest. To cope with this limitation, we propose a time-sensitive and stabilized interest similarity model that involves a process of calculating the similarity of user interest. Moreover, by combining those two kinds of similarity model based on weight factors, we develop a novel algorithm for calculation, which is named as IPSTS (IPSTS). To evaluate the proposed approach, experiments are performed and results indicate that Mean Absolute Difference (MAE) and root mean square error (RMSE) could be significantly reduced, when compared with those of traditional collaborative filtering Algorithms.