Research On Recommendation Algorithm Based On Social Network Analysis

With the development of mobile Internet and the popularity of smartphones,people are enjoying the convenience of technology,and the problem of information overload has become increasingly severe.How to help users to accurately select the information that users are interested in from the vast amount of data is recommended to the user,which is the problem that the recommendation system should solve.Some shopping or review sites such as Amazon and Epinions are rapidly emerging,and the data obtained by these service providers has serious sparseness and fragmentation problems.This leads to a significant reduction in accuracy and recommendation quality based on these sparsely dispersed data,making it impossible for users to obtain satisfactory recommendations.Traditional methods of recommendation,such as collaborative filtering,have some inherent problems such as cold start,low efficiency,and low accuracy.And when the size of users and projects increases,the recommendation performance of collaborative filtering algorithms declines rapidly.Due to the shortcomings of the traditional recommendation algorithm,it is necessary to change the existing recommendation method and introduce a new data source to improve the recommendation performance of the system.On the other hand,the existing recommendation algorithm can only mine based on all historical feedback information of the user to find the user's preference.The feedback information is fused together,and there is no prioritization.The algorithm ignores the user's recent preferences and only considers the user's long-term preference.It is well known that some people's preferences are not static,but will change over time,so mining the short-term preferences of users is especially important for recommendation systems.In view of the shortcomings of traditional recommendation methods and in order to more accurately mine user preferences,this paper mainly studies from the following two aspects:(1)The recommendation algorithm based on social trust clustering is studied.On the one hand,the social trust relationship can reflect the mutual influence of users and common interests.On the other hand,because the traditional recommendation algorithm has a cold start problem,the algorithm that integrates social trust can effectively solve the problem.In this paper,the similarity calculation is performed for all users according to the user-item scoring matrix.At the same time,the user is clustered according to the trust relationship between users,and the membership matrix of the user and the class can be obtained.Then the method of calculating the trust value between users based on the membership matrix is proposed.Finally,the linear combination of user similarity and trust value is recommended.The experimental results on the Douban and Epinions datasets show that compared with the traditional CF-based,trust-based and user-project clustering recommendation algorithms,the algorithm can greatly improve the recommendation quality and improve the time efficiency of the algorithm.(2)The Markov sequence recommendation algorithm discovered by the fusion community is firstly classified into users according to the user's social relationship,and the users classified in one community are regarded as similar users.Users in the same community share a transition probability matrix,and a community's transition probability matrix is calculated,relying on the history of all members in the community.This can not only solve the sparse problem of transfer matrix data between projects,but also improve the time efficiency of the algorithm.In addition,this paper integrates the forgetting attribute into the personalized Markov transition matrix calculation,which can further improve the data sparseness.Since the Markov model can only mine the short-term preferences of users,in order to mine the long-term preferences of users,this paper uses the similarity method of the items.The recommendation result of the personalized Markov model that integrates the user's long-term and short-term preferences will be more in line with the actual needs of the user.The experimental results on the Ciao and Epinions real data sets show that the proposed algorithm can effectively improve the recommendation quality compared with other algorithms.