Research On Recommendation Algorithms Combining Time Effect And User Attributes

With the rapid development of network information technology,the information on the network has achieved leapfrog growth.It takes a lot of time for users to obtain useful information from mass information.Meanwhile,how to screen the content that meets users' interests and hobbies has become an urgent problem for major software and websites to solve.With the rapid development of recommendation system,new recommendation algorithms are constantly put forward and studied.As a traditional algorithm the collaborative filtering algorithm has been researched by many academics,but it also has weak points.Collaborative filtering algorithms only consider user-rating or item-rating information,with a single consideration angle,and user behavior data is too sparse.Failure to consider the impact of user information on the recommendation effect and the changes in user interests over time.In response to the problems mentioned above,this paper has done the following work:1.The traditional similarity only considers the user project rating information,which reduces the accuracy of the user similarity.Therefore,the user attribute information is introduced into the user similarity as the comprehensive user similarity,and It is introduced into the weighted Slope One algorithm to calculate the project Dev to predict the project score,ease data sparseness,and improve the precise of the algorithm.2.In view of the problem that the collaborative filtering algorithm considers a single perspective,this paper introduces the time factor and considers the time factor.The effect of inter-effect on user interest.The user similarity under the time factor was calculated by Ebbinghaus Forgetting Curve,and the prediction score was made according to the improved user similarity and similarity support.Finally,the two algorithms are fused to solve the problem existing in the single algorithm,namely,the unpredictable shortcoming of the Slope One algorithm that the two items have no common score,which improves the limitations of the single algorithm and improves the accuracy of prediction.Finally,this paper applies the algorithm to the Movielens dataset to verify the effectiveness of the fusion time effect and user attribute algorithm.Through experimental comparison and analysis,the improved algorithm has good results on the MAE and RMSE values,which improves the recommendation performance.