Hybrid Recommender System Based On User Profile And Collaborative Filtering

With the rapid development of the Internet,information has grown exponentially.Facing the ocean of information,how to quickly and effectively help users to obtain the information they really need becomes a challenging task and it is also a hot issue in academic research.Academia and the industry have carried out a lot of research and practice on the problem of information overload,and have proposed a variety of solutions for information personalized,hoping to provide users with information that meets their needs.The recommender system is one of the most effective ways to solve this problem.Recommender system which implements personalized recommendation through recommendation algorithms is a personalized information service system.And collaborative filtering is one of the most successful and widely used techniques in the recommender system.Collaborative filtering predicts user's ratings of unknown projects by user's data of the historical ratings of the projects.However,with the continuous development of the Internet,the number of users and projects which on large platforms like TaoBao has reached nearly 100 million levels and is still increasing.The user's rating of the project is only a small part of the total number of items,resulting in extremely sparse of user ratings matrix which is used to predict unknown ratings.And it is difficult to find users with similar preferences for target user by using sparse ratings matrix,then the quality of recommendations produced by ratings matrix is getting worse.The sparsity of the rating matrix has become a key issue affecting recommendations based on collaborative recommendations.Therefor,we needed a new method to solve this problem.User profile can accurately and efficiently analyze the user's preference information,so combining it with collaborative filtering may improve the problem of sparsity.And some surveys show that 80% of users are willing to provide their basic information such as name,age and gender for platform.Therefor,author has researched how toconstruct user profile using user basic information and how to integrate user profile into collaborative filtering.And author proposed a user information measurement model for constructing user portraits,a hybrid recommendation model UPCF based on user profile and collaborative filtering.In the experimental stage,the effects of similarity measurement models(PCC,COS,ADCOS),scoring prediction algorithm(DFM,WS),and user characteristics on the hybrid recommendation model UPCF are investigated,and the optimal combination of UPCF model is selected.Then the UPCF model is compared with UBCF,IBCF and SM models under the four evaluation indexes of MAE,Precision,Recall and F1.The experimental results show that the proposed method is superior to the traditional methods UBCF,IBCF and SM.The result proves that the proposed method improves the prediction accuracy of the collaborative filtering recommendation algorithm and mitigates the impact of data sparsity.