Research On Collaborative Recommendation Algorithm Based On User’s Importance

With the rapid development of electronic commerce, human beings have stepped into network economy era. Actually, it is usually difficult for users to find the most appropriate or needed products from a lot of commodity information. Therefore, the users hope that electronic commerce system can be similar to the purchasing assistant, it can help users select products. Electronic commerce system will automatically give the users a recommendation for the most interested product. In order to solve this arising problem, Recommender System(RS) has came into being, it predicts user’s interest by setting up the connection between the user and product information. On the one hand, RS can help users find the product what they are most interested in. On the other hand, it can also make the product to be showed in the eyes of the people who interested. Thus the users and suppliers can get double-win. Since RS was put forward independent as theory int the mid-1990s, the researchers have came up with many recommendation algorithms that applied to practical application.Although recommendation technology has already been widely applied in the Internet, there are also some general problems needed to be solved, such as data sparse causes the low accuracy and highly vulnerable. Traditional collaborative filtering Recommender System did not consider their relationship among different users. However, different users have different historical behaviors and social relations which lead to various influences. We present two recommendation algorithms based on user’s importance, and user’s importance can be measured from two aspects including user’s social importance and user’s reputation importance respectively. Therefore, the main work in this dissertation is summarized as follows:(1) In this dissertation, we combine the Recommender System’s development process and give detailed introduction to classical collaborative filtering recommendation algorithm. We summarize the problems existing in the current collaborative filtering recommendation algorithm and analyze the related researchers’ work. And two new collaborative recommendation algorithms are put forward based on the importance of user.(2) A recommendation algorithm based on user social relations importance is put forward in this dissertation. Firstly, network topology of user’s social structure and the relationship between the users are analyzed. Secondly, the user’s local importance and global importance are calculated by user similarity and social network respectively. On the one hand, in order to achieve the purpose of improving the recommendation quality, the importance as weights is added into recommendation to adjust the different influence of the users. On the other hand, the global importance of the user is not affected by data sparse, to a certain extent the problem of precision in the data sparse is solved.(3)A robust recommendation algorithm based on user reputation importance is put forward in this dissertation. We firstly use the user’s historical records to obtain the user’s reputation, and build a user’s reputation system. We secondly utilize the Latent Factor Model (LFM) in the field of collaborative filtering recommendation, and finally present a robust collaborative recommendation algorithm based on user’s reputation. Our algorithm improves the system’s robustness from two aspects of shilling attack and natural noise. Experiments on Movielens1M datasets show that comparing with the existed robust recommendation algorithms; this algorithm has the following features:simplicity, interpret ability, stability. And it has strong ability to resist the system attack with the accuracy gets a certain improvement.