Font Size: a A A

Research On Personalized Recommendation Method Of Social Network Based On Power Law And Locality-Sensitive Hash

Posted on:2017-05-16Degree:MasterType:Thesis
Country:ChinaCandidate:B LuoFull Text:PDF
GTID:2428330623454765Subject:Management Science and Engineering
Abstract/Summary:PDF Full Text Request
This paper focuses on the social network research.The main research is several important issues of the social network in the context of the Big Data Age.Social network is the most popular Internet applications and has a large number of users in recent years,such as Facebook,Twitter,Sina microblogging,WeChat moment.All of them have a billion or even billions of user groups and have accumulated a large amount of user behavior data.These people communicate and share personal information through the social network and accumulate a lot of user data for social networks which is of great significance to the research of user behavior,information dissemination,complex network,recommendation system and so on.Research of the data set cover from the management,social science to computer science and technology research.Based on the user data of Sina Weibo.In this paper,Sina microblogging user data on large data background to bring long tail recommendations,data sparsity,data dimension disaster and other issues to carry out research work.Power-law distribution is the basic law of social networks.The long-tail part of power-law data has obvious sparsity.Long-tail recommendation is always the challenge of recommendation system.And cold start,data sparsity and coverage of these issues is also an important part of the recommended system.In this paper,by analyzing the power-law distribution of data and research of personalized recommendation method in social network.Analyzes the power-law distribution of social network user behavior data.The maximum likelihood estimation is used to calculate the scalar value of the power-law distribution.The similarity calculation method is improved by the power law.Finally,a mixed recommendation method based on power law is proposed which is named PowerLawCF(Collaboration Filter).The results show that the recommendation effect of PowerLawCF algorithm is improved significantly,and the effect of long tail recommendation is improved.The data sparsity and cold start problem of recommendation system are better solved.Further more,we study the efficient performance of Locality-Sensitive hash on KNN search,build a suitable hash function family for social network recommendation,and group the users based on hash result,and construct a personalized recommendation method based on Locality-Sensitive hash.The empirical results show that the proposed method has a more effective recommendation effect.This paper solves the long tail recommendation,data sparsity and data dimension disaster in the personalized recommendation of social network by studying the power law of social network,the long tail recommendation method and the application of Locality-Sensitive hash.The value of research is embodied in theory and application.
Keywords/Search Tags:social network, recommendation, collaborative filtering, power law, locality sensitive hash
PDF Full Text Request
Related items