Research Of Collaborative Filtering Algorithm Based On Time Awareness And Social Network Trust

With the rapid development of Internet, the recommender system has been widely used in many areas such as movies, music, e-commerce sites and others. Collaborative filtering recommender system algorithm is the most popular technology, which plays a central role in many areas of the Internet. It can be used to search for similar users or items based on user’s historical rated data, then predict scores. Although the collaborative filtering algorithm has been successfully applied to a large number of e-commerce recommender systems, there are still problems in it such as data sparsity, cold start, user’s interest change and so on. How to effectively solve these issues has become an important area of research. At the same time, a variety of social networking platforms have emerged and are gradually infiltrated into people’s daily lives. Social networks are affecting people’s behavior of choosing and buying items on e-commerce system. It is very significant to re-examine the existing product recommendation system in the new perspective of social networks. According to the important problems above, this paper researches in the following several aspects.First, we research the collaborative filtering algorithm based on the time awareness. Currently, the recommendation system presents a dynamic trend and users’ interest changes over time. Therefore, this paper imports the time information to the collaborative filtering algorithm. At the same time, the traditional collaborative filtering algorithm doesn’t consider the target items when calculating the similarity between users, which results in that the neighbors of a user for all target items are same. Since aiming at a target user for different target items should have different neighbors, this issue presents a new similarity calculation formula. A suite of experiments indicates that the improved algorithm is better than traditional collaborative filtering algorithm in precision, recall, MAE and coverage in these areas.Second, we study the collaborative filtering algorithm on the basis of social network trust. In this paper, to alleviating cold-start problems in traditional collaborative filtering algorithm, we design a new algorithm based social network trust. First, with the help of social network information, we model users. Next, we combine user trust with user preferences to predict the score. Finally, experimental results obtained by two different ways demonstrate that, the proposed algorithm not only enhances the precision of recommendation, but also effectively reduces data sparseness and cold start problem.