Recommender Systems Incorporating Multi-relational Social Trust Networks

With the rapid development of the Internet industry,the amount of data is also exploding in geometric multiples,which lead to information overload,people can't obtain valuable information easily.Traditional search engine technology can deal with the problem of information overload,but it is not easy for search engines to discover users' potential interests from massive data.A recommendation system can solve this problem and meet people's urgent need to obtain personalized information.Unlike a search engine that needs to provide keywords for search,the recommendation system will analyze the user's usage history or historical behavior to discover the user's potential interests and recommend them accordingly.The recommendation system can better meet the user's personalized needs.The recommendation system solves the problem of information overload to a certain extent,but there are problems such as sparse data and cold start.Social trust information has also been gradually applied in recommendation systems with the development of social networks.Research shows that users tend to establish social relationships with people with similar preferences,which is there is a correlation between trust relationships and user similarities,and some studies have shown that users who prefer to adopt recommendations from systems they trust and understand prefer to adopt recommendations from people they know and trust.Therefore,introducing trust relationships into the recommendation system has become a feasible method to solve these problems.However,existing recommendation systems that introduce trust relationships often only introduce a single social trust relationship,resulting in low recommendation accuracy and coverage.This paper proposes a recommendation algorithm that integrates multiple social trust relationships.This algorithm uses a multi-subnet composite complex network model to build a multi-relation social trust network,and introduces multiple social trust information into the recommendation system to improve the accuracy and coverage of recommendations.From the experimental results on Film Trust,it can be seen that the recommendation made on the multi-relational social trust composite network is significantly improved in accuracy and coverage than the recommendation that introduces a single social trust relationship.