Implicit Trust Recommendation Algorithm based on Item Meaning

Abstract

Incorporating user trust relationships can effectively alleviate the cold start and data sparse problems in traditional collaborative filtering algorithms. But in real life, the user trust relationships that can be collected also face data sparse problems. In addition, the existing recommendation algorithm based on user implicit trust relationship does not consider the negative and enthusiasm of certain products that have special significance to users and the common experience items between users when mining implicit trust relationships among users. In response to this type of problem, this article uses the historical scoring information of users to infer the meaning of the project to users, and the influence of the negative and enthusiasm of the user experience project on the implicit trust relationship between users, supplementing the existing trust statement between users to improve the accuracy of recommendation. Firstly, divide the community based on user trust relationships. Second, use the divided community user rating information to dig out the meaning of different items for different users, and infer the implicit direct trust value between users based on the positive and negative ratings between users, then the implicit direct trust between users is used to obtain the implicit indirect trust value according to the trust propagation. Finally, the implicit trust relationship between users was obtained according to the implicit direct and indirect trust between users, and the implicit trust between users was used to replace the similarity in traditional collaborative filtering, so as to find the recommended nearest neighbors of target users and predict the score of missing items. In the experimental results of the real data set FilmTrust and Epinions, compared with the comparison algorithm given in this article, it can effectively improve the accuracy of recommendation.