Research On The Matrix Decomposition Algorithm Based On Time Factor

The improvement of modern network technology brings about the problem of the information overload.It is difficult for people to find the information they need.In order to solve the problem,the recommendation system and a lot of improved extensions have been proposed.Many other models are proposed to solve data sparseness problems,such as matrix decomposition models.They can dig out the behaviors of users from the users and behaviors data set,and give users more accurate recommendations.However,the matrix decomposition model also has some problems.One problem is that the matrix decomposition model only focuses on user's behavior data,but does not consider the change of the user's interest from the angle of time.Although there are many methods to improve the matrix decomposition model by integrating the temporal factors,but the rules of time they used are not been dig out from the real user's behavior data.In the thesis,the main purpose is import the time factor into the matrix decomposition model.First,this thesis surveys a great deal the real user behavior data sets and analyze of the laws of time.Then the regression analysis of the results obtained by statistical analysis,analyzing the time factors how affecting the user interests and item prevalence.Finally,the time model of item popularity and the time model of user interests are integrated with the matrix decomposition model based on bias.This thesis also analyzes the effect of time and item bias on the recommendation results.Finally,it is verified by experiments that the improved matrix decomposition model based on time factor can improve the accuracy of the recommendation system.