Research On Recommendation Algorithm Based On Temporal Effect And Attribute Information

With the rapid development of the Internet,the amount of information has exploded exponentially.Faced with such a huge amount of data,there are some problems in finding the content of the real interest in the ocean of data.The birth of personalized recommendation system is an effective solution to the problem of information overload following the search engine information retrieval technology based on keyword search.It provides personalized information services by mining users' interest preferences,and has more pertinence and improves users' experience.The traditional collaborative filtering algorithm,as the core algorithm of the recommendation system,is easy to understand and has achieved great commercial success.However,there are still some problems: 1)lack of consideration of changes in user interest caused by the time effect,and cannot adapt to the changes of user interest;2)do not consider the participation factor of the user,and item ratio factor;3)the problem of the sparsity of data.These problems restrict the accuracy of recommendation algorithm.Therefore,this thesis improves the traditional collaborative filtering algorithm based on the above problems.In this thesis,the time effect is introduced to the similarity calculation of the recommendation algorithm as an important factor,which will give more weight to the similarity of the users with similar scoring time and age,and do help to build the nearest neighbor set.In addition,the implicit feedback information of the user includes user participation degree and user rating item ratio factor are fused as a user implicit attribute influence factor that is added to similarity calculation to modify the influence of near neighbor.Experiments show that the improved recommendation algorithm can improve the accuracy of prediction.In addition,the sparsity of data is more obvious after the time effect is introduced.In order to alleviate the problem of data sparsity,this thesis uses the user rating information and time effect based on the property information,and extends the score matrix.In this way,data sources are enriched,users' interest preferences are calculated from the dimension of attributes.Finally,the above two improved methods to calculate similarity are linearly fused,and the parameters are determined by experiments.Experimental results show that the recommendation algorithm incorporating time effect and attribute information can improve the accuracy of recommendation.