Research On Collaborative Filter Recommendation Algorithm Based On Data Mining

With the rapid development of Internet technology,"data explosion" has become a hot issue,and it is increasingly difficult to lock in the choices that best meet the standards in their own minds.As an important means to solve this problem,the recommendation system can obtain the user's preferences through the user's historical behavior and social behavior,and find the user's most likely "interested" recommendation to the user in the mass data.However,the problem of sparse and cold start and personalized recommendation can be seen in the prediction based on user history score data.The text studies these two problems and proposes two effective solutions:A personalized recommendation algorithm for association rules based on information entropy is proposed.Firstly,the paper introduces standard deviation value and weights the scoring information of popular items to reduce the influence of too many association rules.By splitting the strong association rules of the Apriori algorithm,the user set similar to the specified user is obtained to predict the score.In the process of forecasting scoring,the idea of information entropy is used to increase the weight of the user's small probability score,so that the score is more in line with the user's preferences,so as to achieve the purpose of personalized scoring.A recommendation algorithm based on clustering and implicit semantic model is proposed.In order to solve the problem of data sparsity,this paper uses implicit semantic model to fill the user-project matrix.Using fuzzy C-clustering to cluster users,a number of user clusters are obtained.When recommending a specified user,first look for the cluster that is most similar to the specified user,and look for the user that is most similar to the specified user in the cluster,using the similarity between the specified user and the specified user as a weight to calculate the specified user's prediction score for the unknown item..At the same time,in order to achieve the purpose of personalized recommendation,the user score bias and user item type preference are proposed,the basic score is obtained by using fuzzy clustering,and the user score bias and user type preference are combined.The multivariate linear regression is used to calculate the specific gravity of each part and get the final score.Experimental results show that the proposed algorithm has better results.In this experiment,RMSE and MAE are selected as the indicators to measure the accuracy of the recommended algorithm,and the recommended algorithm and other recommended algorithms proposed in this paper are tested and compared on the Movielens dataset.The result shows that the recommendation algorithm proposed in this paper is more effective.