Collaborative Filtering Based On Social Networks And Privacy Protection

With the rapid development of the Internet and the continuous introduction of various Internet applications, network application data also explosive growth. Faced with such a large amount of data information, users have been unable to directly obtain the effective information. To address this information overload problem, researchers had proposed a number of solutions,of which the recommender system is representative and achieved better results. Despite the achievements in the field of recommender systems are significant at present, there are still some problems waiting to be solved. Such as the problem of dealing with massive data, cold start problems, the accuracy of the system and so on. On the other hand, the operation of recommender system need to collect as much as possible users’ information and behavioral data, etc. And the leakage of user data events due to the network application loopholes have occurred from time to time. Therefore, how to protect the users’ privacy data security has become a hot research.This paper firstly introduced the research background, basic principle and research hotspot of the recommender system, summarized the commonly used recommendation model and algorithm realization way and the advantages and disadvantages of them. Then, this paper mainly studies in improving the system accuracy when combined with social network analysis and protecting the user privacy in the recommender system. This paper puts forward the corresponding solutions, and the concrete research content is:(1) An improved algorithm is proposed based on social network user information and Research on scoring information. Firstly, this paper presents A social network users similarity scoring algorithm based on algorithm is adopted to build user trust model charge, use of social network users personal interest information, customer relationship data is presented in this paper. Also in case of sparse data, the similarity algorithm, effectively alleviate the sparse data problem, determine target users in nearest neighbor set, according to the formula of social network user similarity score prediction. Then for based on model of collaborative filtering algorithm had a high time-complexity, we had used an improved of collaborative filtering algorithm, solves the matrix with the latent factor model, on the basis of the solved matrix and users’ characteristic data, predicts users’ ratings on items by means of linear weightings, which is generally referred to as the users’ preferential rating in this paper. And finally, It is proposed in this paper an algorithm of inosculating the users’ trustful rating and preferential rating in learning with machine, which uses the user’s trustful rating, preferential rating number of neighbors on the social network and number of characteristics as the training features, and takes the user’s actual rating as the label, then uses the Fisher linear discrimination as the training model to predict user’s rating on certain items.(2) Considering the problem of user’ privacy protection in Social Network, this paper proposes a collaborative filtering recommendation algorithm based on item character information. The algorithm uses the characteristic similarity information to estimate the user’s preference of a specific item, and then it is recommended. Finally, the experiments show that the algorithm in recommendation accuracy has a certain loss, but still within the range of applicability. On the other hand, as the input data of the system is only some generalization of data, user privacy has been effectively protected. So the algorithm has high practical application value.