Research On Service Recommendation Based On User Similarity Relation And Service Association Relation

Nowadays,Web services with diversified functions are everywhere on the Internet,which makes service providers put forward higher requirements on how to dig out Web services that meet users' functional needs in the massive service resource pool.The matrix factorization model based on the fusion of social information has become one of the mainstream methods for service recommendation due to its good prediction accuracy.With the continuous improvement of users' personalized needs,the existing service recommendation methods based on the probability matrix factorization model of integrated social information still have some shortcomings in the accuracy of service recommendation.Firstly,the lack of in-depth exploration and dynamic analysis of users' personalized preferences leads to inaccurate solution of user similarity relations.Secondly,the influence of the association relationship between services on the accuracy of service recommendation is ignored.Thirdly,the original scoring matrix is sparse and the scoring confidence is not high.This paper studies the service recommendation method based on user similarity relationship and service association relationship,aiming to improve the accuracy of service recommendation in the case of data sparsity and cold startup problems.Based on the analysis and summary of the research status of service recommendation,this paper presents a service recommendation framework based on user similarity relationship and service association relationship.Firstly,a user similarity algorithm based on dynamic preference is proposed.This algorithm is aimed at solving the problem of inaccurate calculation of user similarity relationship.The user's rating preference is calculated based on the tags and rating data of service annotation by users.The interest attenuation function is introduced to solve the user's dynamic rating preference for the tag.And the similarity relation between users is calculated based on the dynamic rating preference.Then,the tags importance are calculated based on the idea of tag co-occurrence,and a service association relationship mining algorithm based on weighted bipartite graph is proposed.This algorithm represents the co-occurrence of tags in the form of weighted bipartite graph,and the correlation relation between tags is obtained by bipartite graph resource allocation method,and the correlation relation between services is obtained by combining the tag information of services.Finally,a service recommendation algorithm based on probability matrix factorization model is proposed.The algorithm integrates the user similarity relationship based on dynamic preference and the service association relationship based on weighted bipartite graph mining into the probability matrix factorization model,predicts the user's rating of the unknown service,and finally recommends the Top-N rated service to users.Experiments were carried out on real data sets.RMSE values and MAE values were used as the evaluation criteria for the accuracy of service recommendation.Representative methods were selected for experimental comparison.The experimental results show that the proposed service recommendation method(US-SA-PMF)is superior to other representative service recommendation methods in recommendation accuracy.