Research Of Social Recommendation Algorithm Based On Time Series Model

In recent years,the Internet technology is developing rapidly,and people have entered the era of the big data from the era which lacks information.In such a situation,especially with the development of social network technology,when people choose services on the Internet,they trust and rely on their friends to a higher degree.However,with the emergence of massive information,people's social network relationship is becoming more and more complex,and users can hardly find the potential interests reflected in the social network relationship only with their own eyes or hands.Based on these problems,how to realize the effective use of social network relationship,and how to combine the user's personal information for information recommendation,both become problems in the development of recommendation system.In order to solve the problem of cold start and data sparsity in traditional recommender system,the researchers began to use social recommendations to provide services to users.However,in the face of changing user needs and increasingly complex recommendation tasks,social recommender systems also need to incorporate more data and algorithms to provide better recommendation services.In this paper,on the basis of the existing probability matrix factorization theory,according to the characteristics of the user behavior and the changing interest with time evolution,a recommender algorithm based on time series model and matrix factorization is proposed.Then,this paper constructs a social recommendation algorithm based on time series model that combines the advantages of social recommendation methods.After that,according to this algorithm,this paper designs and implements a social recommendation system prototype.This paper mainly includes the following work:(1)Analyze the research background and research status of recommender systems and social recommendation,according to the characteristics and shortcomings of the main current recommendation algorithm,put forward the main contents and innovations of this study,expound the problems and significance of this research.(2)Research the related technology of time series model and probability matrix factorization,analyze the main theoretical basis of social recommendation,discuss the social recommendation technology based on the collaborative filtering,matrix factorization,and probability model.On this basis,use the user behavior data to analyzethe user interest drift,and propose a recommendation algorithm based on matrix factorization and time series model.Describe the algorithm framework and implementation process,and analyze the time complexity of the algorithm,evaluate the performance of the algorithm by the comparing experimental on the real Alibaba music dataset.(3)Based on the above algorithm,use social regularization methods to improve it,and push model formula of socialization probability matrix factorization.Propose a social recommendation algorithm based on time series model,and compare this algorithm with other social recommendation algorithms on Douban movie and social network dataset.(4)Study on the social recommendation algorithm based on the time series model,design and implementation a social recommendation system prototype,and describe the overall framework of the system and the main function modules in detail.