摘要
协同过滤推荐算法是在电子商务推荐系统中应用最成功的推荐技术之一。提出了一种基于共同评分和相似性权重的协同过滤推荐算法。该算法选择用户的共同评分数据计算用户的相似性,选择项目被用户共同评分的数据计算项目的相似性,再分别计算基于用户以及项目算法的预测评分,然后通过相似性权重结合两者得到最终的预测结果,最后再根据预测结果产生推荐。实际数据的实验结果表明,提出的算法显著提高了预测准确度,从而提高了推荐质量。
Collaborative filtering recommendation algorithm is one of the most successful technologies in the e-com- merce recommendation system. This paper presented a collaborative filtering algorithm based on co-ratings and similarity weight. First, the co-ratings were selected to compute the similarity between users or items. Most importantly, the algo- rithm acquiring the last prediction result was acquired by combining prior predicting rating with similarity weight, from which recommendation was produced. The experimental results in real data show this algorithm can consistently achieve better prediction accuracy, thereby brings better recommendation quality.
出处
《计算机科学》
CSCD
北大核心
2010年第2期99-104,共6页
Computer Science
基金
国家自然科学基金(60573097
60773198
60703111)
广东省自然科学基金(05200302
06104916)
广州市科技计划项目(2007Z3-D3071)
高等学校博士学科点专项科研基金(20050558017)
新世纪优秀人才支持计划(NCET-06-0727)资助
关键词
电子商务
推荐系统
协同过滤
共同评分
相似性权重
E-commerce,Recommendation system,Collaborative filtering,Co-rating,Similarity weight
