The Research And Application Of Group Recommendation Algorithm Based On Matrix Factorization

With the rapid development of Internet technology,the network continues to penetrate into people's lives.Internet users as Internet information consumers,but also Internet content producers.With the increase of network data,the problem of "information overload" is becoming more and more serious.In order to solve the "information overload" problem,find the most valuable information in the massive data quickly,people put forward personalized Recommender System(RS).However,in the past,personalized recommender research focused on individual user,and can not meet the current recommendations of group user activity,therefore,the rise of research for personalized group recommendation.Group recommendation is an extension and expansion of individual recommendation,and because of differences in preferences between users of group,the characteristics of individual users are different,so the groups recommendation more complex than individual recommendation.Therefore,it is of great significance to study the personalized group recommendation for social group activities.Firstly,this thesis makes a deep research on the background of individual recommendation research,the classification of individual recommender system,and the important recommendation methods and key technologies.The traditional individual recommendation method: content-based recommendation algorithm and collaborative filtering-based recommendation algorithm are introduced in detail.Depth study of the difference between group recommendation and individual recommendation problems and the key techniques used in group recommendation research,such as group user preference acquisition,group discovery,preference aggregation strategy and so on,and the group recommendation is faced with new challenges for a comprehensive analysis.Secondly,on the basis of studying matrix factorization model,the thesis mainly introduces the implementation steps and key techniques of group recommendation algorithm based on matrix factorization: random gradient descent method and group aggregation function.The basic matrix factorization model,regularization matrix factorization model and deviation matrix factorization model.The matrix factorization model is improved by ridge regression analysis,and a matrix factorization model(WMF model)is proposed to combine the weights of the items.Considering the change of userpreferences over time,the fitting time function is proposed,and the time effect function after fitting is combined with the WMF model to further improve the algorithm.Finally,the movie group recommender system is designed and implemented.The main functions of the system include constructing groups,generating group recommendation,recommending result display and so on.The movie group recommender system can more intuitively reflect the operation process and practical significance of group recommendation algorithm.