Research On Collaborative Filtering Algorithm Based On Social Network

With the development of the digital information,the online social network service has been developed rapidly.Because of explosive growth of network,it is difficult for users to find the useful information from mass information.Thus,personalized recommendation system is proposed,and the research of recommendation system based on social network has become a hot topic.The structure of social network,geographical location services and a variety of recommendation algorithms were analyzed in this paper.The influence of the factors such as trust relationship,user's rating and geographical position on the personalized recommendation algorithm is studied.Compared with the traditional recommendation model,the recommendation model based on social network has serious problem of data sparseness.In order to solve the problems,two kinds of improved collaborative filtering recommendation algorithm were proposed,the main work and research results are as follows:Focusing on the problem of date sparseness and poor scalability,a new clustering recommendation algorithm based on user interest and social trust is proposed in this paper.Firstly,according to user rating information,the algorithm divides users into different categories by clustering technology,and a user neighbor sets is established by user's interest.At the same time,to measure the implicit trust value among social network users by defining direct trust,indirect trust,transfer path and calculation method,a neighbor set of user trust is established.Finally,two neighbor sets are combined to generate recommendations for users.The simulation experiment is carried out to test the performance on Douban dataset.Compared with other algorithms,the new algorithm has better performance in MAE,precision,recall and F1.Experimental results show that the new algorithm effectively improve the quality of recommendation system.Focusing on the problem of the accuracy of collaborative filtering algorithm,a new location algorithm based on the geographical position social network is proposed,which provide users with location recommendation.Firstly,the potential trust relationship among the users in the social network is excavated,and the trust value among the users is calculated.Then,according to the user's Check-In message,the algorithm analyzes whether there is a common place preference between the users,and the preference value is calculated.Got the position similarity by calculated the distance between the use.Then,in order to predict the location,user interest,social trust,and user distance factor are weighted by alpha and beta.The simulation experiment is carried out to test the performance on Foursquare dataset.Compared with other algorithms,the new algorithm has better performance.Experimental results show that the new algorithm has better performance in MAE,precision,recall and F1.