Collaborative Filtering Recommendation Algorithm Based On Intuitionistic Fuzzy C-means Clustering And User Interest

As the technology of the Internet advances by leaps and bounds,the amount of information and users on the Internet grows exponentially.In the mass of goods,how can users choose,how can sellers provide users with the best products in line with their needs,the emergence of personalized recommendation system has improved this problem.In personalized recommendation system,collaborative filtering recommendation algorithm is the most widely used algorithm,which has relatively good recommendation effect and simple operation,but it will also face a lot of problems.As the scale of recommendation system expands and the number of users and items increases,the collaborative filtering recommendation algorithm has high data sparsity,poor real-time and expansibility of the algorithm,and cold start,which affects the prediction accuracy of the recommendation algorithm.In order to improve the recommendation effect of the collaborative filtering recommendation algorithm,this thesis optimized the collaborative filtering recommendation algorithm from two aspects of real-time performance and data sparsity to improve the prediction accuracy of the recommendation algorithm.The main work is as follows:Firstly,to solve the real-time problem of traditional collaborative filtering recommendation algorithm,intuitive fuzzy c-means clustering algorithm(IFCM)is introduced in this thesis for recommendation.As the key link of collaborative filtering recommendation algorithm,clustering algorithm calculates the nearest neighbor of similarity search object,and introducing clustering into collaborative filtering recommendation algorithm can improve the recommendation performance.As for the classical clustering algorithm,the feature data are accurate and the classification is "either/or".Considering the fact that the user's interest degree is fuzzy,this thesis introduces intuitionistic fuzzy clustering into the recommendation system.It only needs to search the nearest neighbor in the class cluster of the target user,and does not need to traverse the whole data set,which reduces the running time of the algorithm and effectively alleviates the real-time problem of the collaborative filtering recommendation algorithm.Through experiments one and three,the MAE value of the recommendation by introducing intuitive fuzzy c-means clustering algorithm is smaller than that of the traditional CF algorithm,and smaller than that by introducing k-means clustering and fuzzy c-means clustering.Secondly,in view of the data sparsity existing in the algorithm,this thesis introduces user interest to fill in the missing values of the scoring data.The missing data are filled by the mean of the score,which fails to take into account the difference of users' interest in the elements contained in the project,and the mean filling fails to take into account the difference of users,which may affect the recommendation performance of the recommendation system.In this thesis,the user's degree of interest in different elements is obtained through the evaluation table of the project and the table of elements contained in the project,and then missing data is filled in the scoring matrix.While considering the differences among users,the influence of data sparsity on the prediction accuracy of collaborative filtering recommendation algorithm is effectively alleviated.Through experiment two verification,the MAE value of the missing data is less than the MAE value of the mean value of the recommendation after the missing data is filled with user interest.Finally,this thesis combines intuitionistic fuzzy clustering with user interest filling,introduces collaborative filtering recommendation algorithm,and proposes a collaborative filtering recommendation algorithm based on intuitionistic fuzzy c-means clustering and user interest(abbreviated as IFCM-UIR-CF).The algorithm firstly classifies all users using intuitionistic fuzzy clustering algorithm.When the target user selects the nearest neighbor,it only needs to select the nearest neighbor in the class cluster of the target user,without traversing the whole data set.Secondly,in the user class cluster,missing values in the scoring matrix are filled by the user's interest score pairs to reduce the high sparsity of data.Then,the user or item similarity is calculated in the class of the target user,and the first k nearest neighbors of the target object are found according to the order of similarity.Finally,the items of the target user's existing behavior are filtered,and the unknown item score is predicted according to the known user to generate top-n recommendation set to improve the recommendation quality of the algorithm.Through the verification of experiment four,the MAE value of the IFCM-UIR-CF algorithm proposed in this thesis is less than that of the FCM-Slope One-CF algorithm,less than that of the IFCM-Slope One-CF algorithm,and less than that of the traditional CF algorithm.Therefore,the recommendation accuracy is the highest.Four comparative experiments were conducted to verify the effectiveness of the proposed algorithm.The experimental results show that :(1)the prediction accuracy of the IFCM algorithm quoted in this thesis is better than that of k-means clustering for collaborative filtering recommendation,FCM algorithm for collaborative filtering recommendation and traditional CF algorithm.(2)This missing value filling method can effectively reduce data sparsity,and the prediction accuracy is higher than that of mean filling.(3)The IFCM-UIR-CF algorithm proposed in this thesis has higher scoring prediction accuracy and recommendation quality,and is superior to the collaborative filtering recommendation algorithm based on fuzzy c-means clustering and Slope-One,the collaborative filtering recommendation algorithm based on intuitionistic fuzzy C-means clustering and Slope One,and the traditional collaborative filtering recommendation algorithm.In summary: The algorithm proposed in this thesis-a collaborative filtering recommendation algorithm based on intuitionistic fuzzy c-means clustering and user interest has significantly improved the recommendation accuracy and effectively alleviated the problems of data sparsity and scalability.