Research On Collaborative Filtering Recommendation Algorithm Based On Trus

As the Internet gradually penetrated into people’s lives,a large amount of data appeared in all walks of life.Big data brings unprecedented opportunities and challenges for extracting valuable information.Recommender systems are an effective tool to solve information overload.It can mine the user’s preference through the user’s historical data,so as to find the information that the user may be interested in and recommend it.But for huge data and rare interactive information,data sparsity is the most important problem for recommender systems.Data sparsity is one of the main challenges in collaborative filtering.Considering that in reality,when users choose items,they prefer the items recommended by their friends.Then the trust relationship between users provides useful additional information for the recommender systems.The direct trust relationship is mainly used as additional information in some work,but the indirect trust relationship is seldom considered.In view of this situation,two recommendation algorithms combined with user trust relationship are proposed.The details are as follows.ATRec is proposed to relieve the problem of data sparsity.It integrates direct and indirect asymmetric trust relationships.First of all,ATRec constructs a trust transfer mechanism.The mechanism is used to obtain the indirect asymmetric trust relationship between users.Then,each user’s trust set is obtained according to the direct and indirect asymmetric trust relationship.Finally,we use the trust set and the nearest neighbor’s score to calculate the recommendation probability of items,and get the recommendation list for users.Experiments show that can relieve the problem of data sparsity and ensure the precision,but the recall is not very good.ATPMF is proposed to improve ATRec.ATPMF integrates the asymmetric trust relationship of users and the factorization of matrix.Two factor matrices are obtained by matrix factorization,and the optimal solution is obtained by using the least alternate two multiplication.Secondly,a new probability matrix is obtained according to the optimal solution.Then,according to the trust set obtained by the trust relationship,the popularity probability of goods is calculated.Finally,we combine the two probabilities to get the recommendation probability of goods,and then get the recommendation list for users.The experimental results on the real data set are compared,ATRec and ATPMF relieve the problem of data sparsity to some extent.ATRec has a certain improvement in accuracy compared with the mainstream recommendation algorithm.But there are still shortcomings in recall.ATPMF makes up for the shortcomings of ATRec in the recall.