Research On Collaborative Filtering Recommendation Method Based On User Interest Modeling

In the era of Internet,with the rapid development of e-commerce websites and social media platforms,various types of information such as users,products,and comments have exploded.Facing massive and complex information,it is increasingly difficult for users to obtain useful information efficiently,which leads to an increasingly serious problem of information overload.In order to effectively alleviate the problem of information overload,personalized recommendation technology came into being.By analyzing the user's historical behavior information and related preferences,the technology will recommend products,information or services that the user likes to the user.While improving the efficiency of user information acquisition,it enhances user stickiness and promotes relevant stakeholders.Performance improvement.Collaborative filtering recommendation method,as one of the most widely used personalized recommendation techniques,has attracted the attention of academia and industry.Judging from the current research progress,although collaborative filtering recommendation technology has achieved a series of valuable research results,difficulties such as data sparseness and user interest drift still restrict its recommendation performance improvement.This paper faces the problems of data sparseness and user interest drift.Taking the in-depth analysis of user interests as the starting point,by integrating the user's project attribute preferences and interest change information,and comprehensively using nonlinear forgetting functions and kmeans clustering methods,a collaborative filtering recommendation algorithm based on user interest modeling is established.The main research results are as follows:(1)Facing the problem of data sparseness,focusing on more careful consideration of user interests.Based on the fusion of user rating information and item attribute information,this paper transforms the user-item rating matrix into a user-item attribute preference matrix,which reduces the sparseness of the matrix.(2)For the user interest drift problem,based on the Ebbinghaus forgetting curve as the theoretical basis,this paper introduces nonlinear forgetting functions of different ages to describe the user's interest changes in more detail,providing support for optimizing the similarity calculation of the nearest neighbor set.(3)Based on the user's item attribute preference vector,the k-means clustering algorithm is used to classify the users,and the nearest neighbor set acquisition method is given by searching the nearest neighbors of the target user within the class,and then the recommendation generation criterion of this paper is proposed.On this basis,this paper randomly selected 11249 scores of 1682 movies from 100 users from the Movie Lens website ml-100 k public data set as the experimental data basis,and verified the effectiveness and stability of the method in this paper through 6 sets of comparative experiments.In general,this article focuses on the more detailed characterization of user interests and conducts research on collaborative filtering recommendation methods.The research results have a certain theoretical value for dealing with the problem of data sparseness and user interest drift in the recommendation system,and have good application value in the field of personalized recommendation such as e-commerce websites.