Research On Personalized Service Recommendation Method Based On Collaborative Filtering Algorithm

With the rapid development of Web services and related technologies,e-commerce platforms continue to grow,people often consume too much time in their lives to obtain the content they need from large databases.At the same time,people inevitably face a lot of irrelevant data,that is,"information overload" problem.The recommendation system can provide users with personalized service recommendations based on user data records and personal characteristics.Collaborative Filtering(CF)is a more successful algorithm in the existing recommendation system.However,in the case of sparse user rating data,cold start,and user interest transfer,the collaborative filtering algorithm's recommendation ability is affected.In order to improve the recommendation ability of collaborative filtering algorithm,this article has contributions to the following work:(1)Collaborative filtering algorithm based on optimized similarity measure(OSMCF)In view of the deviation of recommendation results caused by the sparse user rating data,this paper proposes a collaborative filtering algorithm based on optimized similarity measures(OSMCF).In the first step,the KNN algorithm is used to initially determine the nearest neighbors for the user and the project.In the second step,the similarity between the two is obtained through the optimized similarity method to further optimize the nearest neighbor.In the third step,the target user is recommended project.The experiment proves that the OSMCF algorithm effectively solves the problem of data sparseness,and the accuracy obtained by the OSMCF algorithm is better.(2)Collaborative filtering algorithm based on user interest transfer(UICCF)In view of the inaccuracy of recommendation items in the recommendation system caused by the transfer of user interest,this paper proposes a collaborative filtering algorithm based on user interest transfer(UICCF)based on the OSMCF algorithm.This algorithm introduces a time penalty function in the process of obtaining the user's rating value for the item.Experimental results show that the recommendation results of the improved algorithm are more in line with user needs.(3)Collaborative filtering algorithm based on item ratings and user characteristics(IRUCCF)In response to the cold start problem of the existing collaborative filtering algorithm,this paper proposes a collaborative filtering algorithm(IRUCCF)based onitem scoring and user characteristics based on the UICCF algorithm.The IRUCCF algorithm combines the scores and user characteristics of users for the same project to linearly merge the project-based CF and the user-based CF.Validated on Movie Lens and Douban datasets,the IRUCCF algorithm improves the accuracy of recommendation results more effectively than other algorithms.