Research On Collaborative Filtering Recommendation Algorithm Based On Item Similarity And User Trust

With the rapid development of Internet technology,it has become increasingly difficult to find information that is really useful to us from massive data,that is,users are currently facing serious information overload problems.The recommendation algorithm can analyze user historical data,analyze the hidden preferences of users,and provide users with information that suits their tastes.It has proved to be an effective means to solve information overload.In recent years,it has received extensive attention and research from all walks of life,and has obtained a lot of research results.However,there are still some problems in the recommendation algorithm,such as the difficulty of coexisting recommendation accuracy and diversity,data sparseness,cold start and malicious users.In this paper,the following research work is carried out for these problems:(1)A matrix decomposition collaborative filtering model CI-MDCF based on auxiliary information is proposed.Based on the traditional matrix decomposition collaborative filtering recommendation model,the model adds user and item auxiliary information to improve the recommendation performance of the traditional recommendation model.(2)A matrix decomposition collaborative filtering recommendation algorithm based on item similarity and user trust is proposed.The algorithm embodies the auxiliary information based on the CI-MDCF model,the article side auxiliary information uses the item attribute similarity information,proposes the item similarity index COS to quantify,and the user side auxiliary information uses the trust relationship information between users to propose the trust degree index CUT to quantify.COS and CUT implement the model to solve the problem that the accuracy and diversity of the recommendation process are difficult to coexist.(3)Propose a trust update algorithm.In the matrix decomposition collaborative filtering recommendation algorithm based on item similarity and user trust,the trust update algorithm is added for the malicious user problem,the trust factor is defined to define the positive behavior and the malicious behavior,and the reward and punishment function is set to trust the user for different behaviors.Update to address the impact of malicious users and malicious ratings on recommendations during the referral process.In this paper,the performance of the proposed algorithm is comprehensively evaluated by comparing experiments on real data sets.The experimental results show that the algorithm can balance the accuracy and diversity of recommendation,and can effectively alleviate the problem of data sparseness and cold start.Users also have a significant effect on the impact of recommendations.