Research On Personalized Recommendation Strategies Based On The Interaction Between Users And Commodities

Personalized recommendation is personalized result for user by quantifying us-er’s preferences for different project,based on the same or related information of user and project.At present,it has been studied extensively at home and aboard,however,there are still many problems and challenges.The accuracy and efficiency of prediction are two concerned issues among them,which is the main content of this paper.This paper builds a cumulative logistic regression model based on the scoring matrix.During the process of building the model,we fully considered the interaction between the user and the project,and we are about to introduce an interactive matrix about them.The number of rows and columns of the interaction matrix would be determined by the number of variables associated with the user and the project respectively.In many practical problems,there are a amount of variables related to the project,for example,we can extract plenty of features to describe a movie in the movie recommendation system which will cause the existences of many unknown parameters in the interaction matrix.Therefore,this paper introduce interaction matrix further into low rank and sparse assumptions.It takes alternating least squares method to estimate the parameters under the low rank hypothesis.And this paper will use the TLP method and DC decomposition to solve the parameter estimation problem in L0 method under the sparse assumption.Eventually,we apply the method proposed in this paper to two movie data of MovieLens.The results show that this method has a good prediction effect.