基于用户行为与角色的协同过滤推荐算法  

Collaborative Filtering Recommendation Algorithm Based on Users' Behavior and Roles

作　　者：李幼平[1,2] 尹柱平[2] 

机构地区：[1]桂林航天工业高等专科学校党委,桂林541004 [2]桂林电子科技大学商学院,桂林541004

出　　处：《计算机系统应用》2011年第11期103-106,共4页Computer Systems & Applications

摘　　要：针对传统协同过滤推荐算法中以稀疏评分计算用户相似性可能并不准确的问题,提出以用户行为对应一定分值填补空缺的I-U评分矩阵,并以分角色下的权重系数K约束用户相似性计算的改进协同过滤推荐算法。实验表明,改进算法的推荐质量更高。There are sparse ratings problem in the traditional CF recommendation algorithm, and based on this sparse ratings will lead to the fact that the similarity may not be accurate. For this reason, a CF algorithm based on fixed I-U ratings matrix, which is given by a certain ratings of user behavior instead of vacancies rating, and weighted coefficient K bases on users＇ role to constrain the similarity calculation is proposed. Experiments show that the improved algorithm has better recommendation quality.

关 键 词：协同过滤 I-U评分矩阵 相似性计算 用户行为 用户角色 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]