User-based Dynamic Collaborative Filtering

With the popularization of the Internet and continuous updating and iteration of computer technology,huge amounts of information are flooded with people's eyes.How to quickly and conveniently find the information they need in the ocean of information? The problem of "information overload" has become a problem that people have to face.Under such circumstances,the recommendation system came into being,and the collaborative filtering algorithm is the oldest and most widely used most successful algorithm in the recommended system algorithms.In this paper,we propose a dynamic collaborative filtering algorithm based on users,which is based on the static,sparse and unicity of users' collaborative filtering algorithm.The main research contents are summarized as follows:For the static nature,introduce the time memory function and establish the dynamic model of user interest.The user's interest does not last forever;it changes over time,so there is some bias in predicting the accuracy of user interest.The introduction of time memory function,fully consider the user interest in different time nodes,the establishment of dynamic models of user interest in order to improve the accuracy of user interest prediction.Aiming at the sparsity of user-item scoring matrix,we modify it by improved item similarity.This article through the improved project similarity,predict the user does not score the project,the score of the project,This effectively avoids the error of user 's similarity caused by the sparsity of user-item scoring matrix.Aim at the unity of item similarity measure,the similarity of project features is introduced,and the similarity of traditional items is improved to improve the similarity of traditional items.The traditional project-based collaborative filtering in the calculation of project similarity,just simply consider the user's historical score of the project,did not consider the project's own attributes,and the similarity of the two projects is essentially the nature of the two projects,it is Not to the user's will for the transfer,is the objective existence of the internal factors.Experimental results show that the improved user-based dynamic collaborative filtering algorithm greatly improves the recommended accuracy.