Research On Enhanced Social Recommendation Based On Comprehensive Trust

With the rapid growth of today's social Internet technology,social recommender [1]has emerged as a major research hotspot in the recommendation systems [2].This paper uses user's comprehensive trust relationship as a supplement,combined with deep learning technology while taking into account the time factor to study social recommendation systems.Based on the Diff Net [3] method,an improved and enhanced social recommendation model,Enhanced Trust Diffusion Neural Network(ETDNet)based on comprehensive trust is constructed.Through analysis and research on the Diff Net recommendation model,the model has the following problems:(1)trust problem: traditional recommendation often lacks the ability to distinguish users' trust degree.In reality,users may trust their friends' feedbacks rather than the ordinary when making decisions.When users interact with different items,users have different levels of trust towards different users.(2)time problem: there is a problem that time influences the inference of users' preference,which has not been well studied in Diff Net.(3)noise problem: Diff Net does not consider the impact of noise generated by long-distance social relations on recommendation performance.Aiming at the mentioned problems of the Diff Net model,an enhanced social recommendation method based on the comprehensive trust relations and taking into account time factors is proposed to further improve the performance of social recommendation system.Firstly,the comprehensive trust based on the historical interaction records of users and items are integrated into the recursive social dynamic modeling to obtain the comprehensive trust of different users toward different items.Secondly,social trust information is captured based on the attention network mechanism as a part of the factors that affect user preferences,so as to solve the problem of weight distribution in the same level domain,while taking into account the time factor.Then,considering the network structure of the model,an optimization method of creating residual connections [4] is designed to reduce the noise of long-distance relations.Finally,the above three parts are merged into a unified framework to enhance each other and construct a better extension model.Through the analysis of experimental data,results on two real social datasets reveal that the ETDNet is superior to other baselines,which improves recommendation results and optimizes recommendation performance,and can provide valuable suggestions and experience for social recommendation.