Research On Collaborative Filtering Recommendation Algorithm Based On User 's Synthetical Interest

The era of big data,Internet information and resources have been serious overload,the quantity of the goods and the users on the electronic commerce website is growing every day,the time and the energy people spend on the process of selection of a good more and more.Search engine and recommendation technology emerge as the times require,and the latter can provide personalized services and can accurately match the needs of users,becoming the focus of current research.Since 1990 s when the recommendation system was born,collaborative filtering algorithm is studied by many scholars of all ages,and has also play a role in practical application,such as the Amazon website.Amazon website's recommendation system is using collaborative filtering technology,and there are about 35% of the sales in relation to it.With the increasing complexity of network information,the shortcomings of collaborative filtering technology are also revealed.From user's interest point of view,the user's behavior is analyzed and the information of user's behavior is incorporated to the recommendation process of collaborative filtering algorithm.Firstly the user's interest are divided into three categories: the user's objective interest,the user's subjective interest and the user's synthetical interest;then based on the user's objective interest model and the user's subjective interest model to improve the traditional collaborative filtering algorithm: 1)Collaborative filtering recommendation algorithm based on user's objective interest also called OICF algorithm is proposed.In the OICF algorithm,the user's objective interestingness is obtained by the fuzzy C mean clustering technology,which is used to fill score matrix.The algorithm improves the traditional item based collaborative filtering algorithm,alleviating the problem of inaccurate recommendation due to the sparsity data;2)Collaborative filtering recommendation algorithm based on user's subjective interest also called SICF algorithm is proposed.In the SICF algorithm,a subjective interestingness matrix based on rating is constructed through the statistical method instead of scoring matrix calculating similarity to improve the traditional user based collaborative filtering algorithm.The algorithm relieves the problem that the number of the score matrix are too much to calculate due to the number of users and items increasing;3)The two kinds of improved recommendation results are combined in UICF algorithm,that is the collaborative filtering algorithm based on synthetical user's interest.The recommendation accuracy is improved,and each improvement effect is verified in the movielens 100 k data set;Finally,according to the basic process of data mining process model of CRISP-DM with the Book-crossing data set test and verify the improved algorithm once again.