基于联合相似用户-项目的协同过滤推荐算法  

Collaborative Filtering Recommendation Based on Cooperative User-item Similarity

作　　者：郑荔平[1] 杨红和[1] 童先群[1] 陈永恒[1] 

机构地区：[1]闽南师范大学计算机学院,福建漳州363000

出　　处：《闽南师范大学学报（自然科学版）》2016年第4期12-21,共10页Journal of Minnan Normal University：Natural Science

基　　金：国家自然科学基金(61303131);福建省教育厅科技项目(JA14192)

摘　　要：由于用户评分的偏好性,及其稀疏的评分矩阵,导致对目标用户的近邻无法进行准确的搜索,使得推荐结果不尽如人意.本文提出了一种联合用户-项目的推荐算法,不仅考虑用户近邻的推荐作用,也考虑了项目近邻的推荐作用.首先,定义相似用户评分影响为用户偏好影响因子,定义相似项目影响为项目偏好影响因子.其次,联合用户偏好影响因子和项目偏好影响因子进行共同推荐.在此基础上,设计新的推荐模型.通过大量的实验证明,提出的新模型得到的推荐质量优于传统模型.Aimed at user-based collaborative filtering, the nearest neighbors of target user are restricted by the extreme sparsity of user＇s rating and the tendency of user＇s rating. In this paper, collaborative filtering recommendation based on cooperative user-item similarity is presented. The recommendation algorithm not only eontains the effect of nearest user similarity, but also the nearest item similarity. Fil^fly,we define the similar user＇sinfluence as the impact factorof user preference,and then define the similar iteminfluenee as the impact factor of item preference.Secondly, we joint the impact factorof user preference and the impact factor of item preference to recommendation. Finally, we design the new recommendation algorithms. The experimental results show that the proposed algorithmcan improve recommendation qualityto a certain extent.

关 键 词：相似性 联合推荐 稀疏性 协同过滤 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]