Collaborative Filtering Recommendation Algorithm For Converting Project Attribute Similarity And Score Similarity

The growing maturity of Internet technology and the rapid growth of Internet users,making more and more industries and enterprises to join the Internet,leading to the rapid growth of Internet information.In face of more and more complex information,the user can not get information from the massive selection of information to meet their own needs timely and precisely,which is called "Information Overload".In order to solve this problem,a variety of personalized recommendation system came into being,which applied the most successful and mature technology is collaborative filtering.But the growing number of users and the number of projects,as well as the effectiveness and timeliness of information on the collaborative filtering recommendation algorithm made a more severe test.This paper focuses on the collaborative filtering recommendation algorithm,which is based on the traditional cooperative filtering recommendation algorithm.In view of the problem that the data sparse and low prevalence projects are recommended in the traditional collaborative filtering recommendation algorithm,this paper proposes a fusion project Attribute Similarity and Score Similarity Collaborative Filtering Recommendation Algorithm.The main contents of this paper are as follows:Firstly,aiming at the data sparsity problem of the user-project scoring matrix in the traditional cooperative filtering recommendation algorithm,considering the influence of project attributes on project similarity calculation,a method of calculating the similarity of weighted Jaccard coefficients of fusion information entropy is proposed,which can be effective Alleviate the sparse data,improve the accuracy of similarity calculation.Secondly,according to the problem of poor accuracy of similarity calculation in traditional collaborative filtering recommendation algorithm and low recommendation rate of low prevalence project,considering the influence of user activity and project popularity on project similarity calculation,the correlation penalty factor is introduced.A method of calculating the similarity degree of the user's degree of activity and project popularity is to improve the recommended rate of the low prevalence project in the case of ensuring the recommended accuracy.Thirdly,in this paper,a collaborative filtering recommendation algorithm is proposed,which combines the similarity degree matrix and the score similarity matrix,to obtain a collaborative algorithm of the similarity degree and the similarity degree of the project.In this paper,The comprehensive similarity matrix,using the TOP-N recommendation,obtains the nearest neighbor of the project and feeds the recommendation back to the user.Finally,the experimental results of the MovieLens dataset are selected and compared with the traditional cooperative filtering algorithm.