Research On Personalized Recommendation Algorithm Based On Clustering

With the development of network resources,personalized recommendation system has become an important tool for network resource query.On the one hand,it can help users save time spent on resource search.On the other hand,it can make the users achieve a satisfactory network resource search even if it has low participation.Personalized recommendation system,the current researching focus,of which domestic and foreign scholars have carried out a lot of research and made great progress.However,there are still many problems.This paper solved the problem of cold start and low accuracy in personalized recommendation system.Also,it analyzed and compared the current advantages and disadvantages of personalized recommendation algorithm.To achieve a reasonable prediction of new user behavior preferences,we make use of the basic attribute element of users is analyzed by using large data.and a user-based MI(Multiple Instance)clustering algorithm is designed.A method of calculating the similarity degree of the weighted sum of the similarity of the user's feature,the basic feature of the project and the similarity of the project score.Based on the subjective-objective deviation,the weighting factor distribution method is designed,The experiment proves its effectiveness and superiority to alleviate the cold start problem and improve the recommendation accuracy.For data sparsity,similar users are clustered through user information feature system.It provides an effective and credible range of calculation for the subsequent scoring data.And then replace the defect value with the statistical average of the item scoring data in the cluster.The final experiment shows the effectiveness of this method in solving the problem of data sparsity.The experimental data set used in this paper is MovieLens-ml-100k,the data set includes training sets and testing sets.At the end of this paper,we use the data set to analyze the algorithm mentioned in this paper,which verified the correctness and superiority of the algorithm.