Research On Key Technology Of Community Recommendation System Based On User Activity And Influence

With the popularity of Internet,Web pages are increasing exponentially and deliver huge information to users.It becomes difficult to find interesting contents in mass information,and there are some "latent information".Traditional search technology can help users retrieve information,however results are usually similar.Personalized recommendation technology relies on users' historical behaviors to analyze their preferences for filtering information.Recently,the academia has attracted more attention to social networks,such as micro-blog,renren,douban.People could share information through social networks and also participate in communities to communicate effectively.It's necessary for websites to adopt corresponding algorithm to increse users' favorability and dependence on them.However,the complexity,asymmetry and overlapping of networks bring challenges to community discovery.How to optimize it has become an urgent problem at present.This paper improves related algorithms in community recommendation.The main achievements are as follows:1.Analyzes and extracts user information in social network platform,and integrates it into community recommendation system.2.A User-Behavior Rank algorithm(UBR)based on users' behavior is proposed,which aims to analyze user influence in social network before recommendation algorithm is executed.UBR algorithm makes recommended results more novel and more realistic.3.Quantify the relationship between users by improving GN(Girvan Newman)algorithm in micro-blog network.The improved GN algorithm is suitable for asymmetric network,and results of community division are obtained by deleting edges of low connection value.4.An algorithm is proposed in combination with ant colony foraging model and signal transmission mechanism.Pheromones will be volatilized at a certain probability,and signal loss mechanism is applied to above process.The application of this algorithm to division of overlapping communities will get better results.5.A CC(Communication Community)community recommendation system is implemented,and UBR algorithm and community discovery algorithm are applied to system.Recommendation system is more pertinent by analyzing users' influence.Meanwhile,our improved community discovery algorithm makes divided results more precise.