摘要
针对个性化推荐系统中协同过滤算法面临的数据稀疏问题以及用户相似性度量的不准确,提出了一种结合类别信息的协同过滤推荐算法。该算法利用用户评分数据计算用户之间对类别关注的相似性,并将用户对类别关注的相似性和用户评分相似性进行组合,得到用户综合相似性,从而提高了最近邻居搜索的准确度,缓解了数据稀疏性问题。实验结果表明,该方法能够有效地避免传统相似性度量方法存在的问题,使得数据稀疏性对最终推荐结果的负面影响变小,在一定程度上提高系统的推荐精度。
Aiming at the difficulty of data sparsity and inaccurate user similarity in personalized recommendation systems, a new algorithm of collaborative filtering using item category information was proposed. The algorithm used user rating data to calculate category concern similarity between users. Category concern similarity and user rating similarity had been synthe- sized to get synthetic user similarity, thus the accurate degree of searching nearest neighbor users has been improved and the sparse of rating data problem has been alleviated simultaneously. The experiment shows that the measure can avoid the defects of traditional similarity measure and reduce the negative effect on the final recommendation and provide better reco- mmendation results for the system.
出处
《重庆邮电大学学报（自然科学版）》
北大核心
2010年第6期823-827,共5页
Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)
关键词
协同过滤
推荐系统
用户相似性
平均绝对误差
collaborative filtering
recommendation system
user similarity
mean absolute error
