Research On The Recommender System Of Matrix Factorization For Social Networking

The rapid development of mobile Internet technology,while improving the quality of life at the same time,generates massive amounts of data.It becomes more and more difficult to find interesting things and meet the demand of people at massive information using traditional search engine.The recommender system can provide people with personalized service to help people find interesting things quickly.In the daily behavior of people's decision making,the surrounding social networking plays an important role.The final judgments of people are the comprehension of themselves and surrounding friends.Matrix factorization algorithm is widely applied in the field of recommender system,which has the advantage of integrating additional information.In order to combine the factor of social networking and advantage of matrix factorization algorithm,matrix factorization algorithm for social networking model(MFSN)is proposed.The model introduces the social networking factor which breaks down the user interest biases as user inner biases and user social biases.Conducting experiments at the crawled Douban dataset and open Foursquare dataset respectively,which adopts cross-validation method and root mean square error(RMSE)and mean absolute error(MAE)measurement.The experiments show that matrix factorization algorithm for social networking is superior than both traditional matrix factorization algorithm and user-based collaborative filter algorithm.The performance of MFSN would decrease first and then increase with the increasing of social networking factor.