Research On The Recommendation Strategy Of Dynamic Context Collaborative Filtering Based On Clustering

With the continuous improvement in China's social and economic fields,the Internet technology has matured and developed rapidly.The recommendation system has also been widely used in various e-commerce websites.The extensive application of the recommendation system can promote the users to browse from the Internet applications The rapid transition of identity to consumers allows users to create more intelligent and efficient information life while bringing more business opportunities to service providers.Collaborative Filtering,as the main technology of recommendation system,can effectively solve the problem of information overload.The core idea of collaborative filtering algorithm is that based on the central idea that users are more willing to adopt the opinions of their friends with the same or similar interests and interests,the nearest neighbor similarity technique is used to study the user's interest characteristics.By predicting several user's points of interest,Personalized recommendations.In the project-user matrix,the number of ratings given by the user is very sparse compared with the number of users,resulting in a sharp decline in the recommended success rate,so the user experience is not ideal because of this problem.In addition,the collaborative filtering algorithm still has some problems,such as cold start problem,not considering the dynamic change of user interests and poor scalability,which leads to the low accuracy of the recommended result.Therefore,we must further study the many problems mentioned above.The specific work of this paper is as follows:(1)It mainly introduces the research background and significance of the research topics,systematically discusses related technologies in recommended fields,including model-based and neighborhood-based collaborative filtering algorithms,systematically summarizes the main data sets in the recommended fields and several commonly used evaluations index.(2)User-based collaborative filtering recommendation method In the process of recommending,it focuses on how to use the historical rating data given by the user to calculate the user similarity.In view of the data sparseness and dynamic scenarios of the traditional collaborative filtering recommendation algorithm,A new dynamic collaborative filtering algorithm based on a weighted clustering method(WCM-DCF)is proposed based on weighted clustering.This method gives more attention to the users who provide more scores.Introducing the concept of user weight based on the SK-means clustering method effectively solves the problem of data sparsity,on the basis of which,it considers incremental updating In order to handle the impact of frequent changes in data during the referral process and to optimize preference predictions and personalized recommendations for target users.The experimental results show that the proposed algorithm can effectively improve the sparseness of user scoring data and provide high-quality recommendation at very low computational cost compared with IUCF,IICF and COCLUST algorithms.(3)Aiming at the shortcomings of traditional text clustering methods in clustering performance,this paper proposes a clustering algorithm based on maximum entropy principle.This algorithm adopts the method of reference cosine similarity in the traditional text clustering algorithm SP-Kmeans,introduces maximum entropy theory to construct the maximum entropy objective function suitable for text clustering,and then introduces maximum entropy principle into spherical K-means text clustering algorithm.Experimental results show that,compared with DA-VMFS and SP-Kmeans algorithm,the performance of CAMEP clustering algorithm proposed in this paper has greatly improved,overall performance is good,which can improve the accuracy and recommended accuracy of clustering.