摘要
提出了利用领域知识进行相似度计算的协同过滤算法,使用户在评分的共同项目很少或为零的情况下也能找到最近邻进行协同推荐。实验结果表明,该算法解决了传统协同过滤算法中相似性度量方法“过严”的问题,在过滤初期显著地提高了推荐质量。
A novel method which exploring hierarchical domain knowledge to computer the user similarity is proposed. This method can find user＇s neighborhood and get predictions by neighbor＇s rating even they have no common rating items. The experiment results show that it can provide better prediction results than traditional collaborative algorithms at such condition.
出处
《计算机工程》
CAS
CSCD
北大核心
2005年第21期7-9,33,共4页
Computer Engineering
基金
国家"863"计划基金资助项目(2002AA142110)
关键词
个性化
推荐系统
协同过滤
领域知识
平均绝对误差
Personalization
Recommendation system
Collaborative filtering
Domain knowledge
MAE
