摘要
针对CF推荐技术依赖的评分矩阵在现实中存在的稀疏性问题,提出用户-项目平均相似度协同过滤推荐算法(ASUCF)。对评分矩阵进行充分挖掘、多次利用,引入平均相似度来惩罚用户或项目的评分或被评分的波动;综合考虑用户和项目两方面,提高预测评分的可靠性。实验结果表明,该方法可以有效提高预测的准确性及推荐质量。
In the user-item rating matrix which is relied on by collaborative filtering,there exists the problem of data sparsity.For this problem,a kind of improved model called ASUCF was proposed.The matrix was sufficiently exploited and repeatedly used.The average similarity was used to punish the fluctuations of user’s ratings or item’s score,and the reliability of prediction score from users as well as items was improved.Finally,the experiments prove that the algorithm can effectively improve the accuracy of prediction and recommendation quality.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第12期4217-4222,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(31301691)
江苏省高等教育教改研究基金项目(2013JSJG195)
关键词
推荐系统
协同过滤
平均相似度
平均绝对偏差
个性化推荐
recommender system
collaborative filtering
average similarity
MAE
personalized recommendation
