摘要
为了提高推荐算法的准确性,充分发挥推荐系统在实际应用中的重要作用,提出了一种基于级联二部图的推荐方法,很好地刻画了在推荐过程中用户和物品之间的复杂关系。在此基础上,充分分析了时间因素在推荐系统中所起的重要作用,将时间属性加入到级联二部图的推荐算法中,进行动态协同过滤的Top-N推荐。基于CiteULike论文数据集的实验结果表明,该方法有效地提高了推荐的准确度,表明了时间因素在推荐算法的研究中是不容忽视的。
To improve the accuracy of the recommend algorithm, and to apply the recommend system into practice, a new recommend algorithm based on the cascaded bi-graph is put forward. The relationship between users and items commendably is described. On this basis, the time factor in the recommend system is analyzed and added in the recommend algorithm based on the cascaded bi-graph to do the dynamic collaborative filtering Top-n recommendation. The performed experiment based on the dataset of CiteULike shows that the efficiency of this algorithm and proves the importance of the time factor in recommend system.
出处
《计算机工程与设计》
CSCD
北大核心
2013年第12期4356-4361,共6页
Computer Engineering and Design
关键词
推荐系统
协同过滤
图模型
动态推荐
时间效应
recommend system
collaborative filtering
graph model
dynamic recommendation
time effect
