With the explosive growth of information in the Internet era,search engines have been unable to meet the needs of users.By analyzing the user’s historical behavior,the recommendation system can predict the user’s preference and realize the active personalized recommendation.However,there are still some problems in the current recommendation algorithm,such as data sparsity and cold start,which affect the performance of the recommendation system.There are user behavior information and user social relations in social networks.Considering the application of social networks to recommendation algorithms,it can effectively improve the above problems.Therefore,this paper first extracts the characteristics of user behavior information in social networks,and then proposes a matrix decomposition recommendation algorithm in social networks.The main research contents are as follows:1.Aiming at the problem that the traditional matrix decomposition algorithm is not easy to integrate multiple features,the model of extracting user preference features is proposed according to the data features in social networks.Using user social relations and user behavior information in social networks to mine user behavior characteristics,from the aspects of social network topology and user keywords,information fusion processing and dynamic reconstruction of user behavior characteristics are carried out to achieve the acquisition of user potential preference characteristics,so as to complete the recommendation in social networks.2.With the development of social networks,it has become a common method to improve the recommendation system by using social networks.However,the information of individual users in social networks is very small,and the problem of sparse data may affect the recommendation accuracy.This paper combines adaptive social network with low-rank objective function,considering not only the relationship between users in social network,but also the comprehensive influence of users and items,and proposes an improved matrix factorization model of stochastic gradient descent.On Epinions and CIAO data sets,the SNMF algorithm proposed in this paper is decreased on RMSE,and embodies better recommendation accuracy.Among them,on the CIAO data set,the recommendation algorithm proposed in this paper improves the prediction error by 9.63%,and on the Epinions data set,the prediction result of the recommendation algorithm proposed in this paper improves by 7.9%.Figure[25]Table[3]Reference[66]... |