《计算机应用研究》|Application Research of Computers

基于MMTD和兴趣偏向系数的协同过滤推荐算法

Collaborative filtering recommendation algorithm based on MMTD and interest bias coefficient

免费全文下载 (已被下载 次)  
获取PDF全文
作者 陆荣,周宁宁
机构 南京邮电大学 计算机学院,南京 210023
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2020)09-007-2600-04
DOI 10.19734/j.issn.1001-3695.2019.05.0108
摘要 针对传统基于用户的协同过滤推荐算法存在的相似性度量不准确和缺乏对用户评分合理应用的问题,提出了一种结合中介真值程度度量(MMTD)和兴趣偏向系数的推荐算法。该方法首先采用MMTD度量用户评分的相似性;然后利用用户评分相似性改进余弦相似性公式和Jaccard公式,得到新的基于MMTD的用户相似性度量方法;最后结合兴趣偏向系数输出推荐结果。在MovieLens-100k数据集上的实验结果表明,该方法可以在一定程度上提高用户间相似性度量的准确性,提高推荐结果的准确率和召回率。
关键词 协同过滤; 用户评分; 用户相似度; 中介真值程度度量; 兴趣偏向系数
基金项目 智能电网保护和运行控制国家重点实验室开放课题(20169,201610)
国家自然科学基金资助项目(61170322,61373065,61302157)
本文URL http://www.arocmag.com/article/01-2020-09-007.html
英文标题 Collaborative filtering recommendation algorithm based on MMTD and interest bias coefficient
作者英文名 Lu Rong, Zhou Ningning
机构英文名 School of Computer,Nanjing University of Posts & Telecommunications,Nanjing 210023,China
英文摘要 Due to the inaccuracy of the traditional user-based collaborative filtering recommendation algorithm and the lack of reasonable application of user ratings, this paper proposed a recommendation algorithm combining the measure of medium truth degree(MMTD) and the interest bias coefficient. The method firstly used MMTD to measure the similarity of user ratings. Secondly, it improved the cosine similarity formula and Jaccard formula by adopting user score similarity, and obtained a new MMTD-based user similarity measurement method. Finally, combining the interest bias coefficient, it output the recommendation result. The experimental results on the MovieLens-100k dataset show that the algorithm can improve the accuracy of the similarity measure between users to some extent, and improve the precision and recall of recommendation results.
英文关键词 collaborative filtering; user ratings; user similarity; measure of medium truth degree; interest bias factor
参考文献 查看稿件参考文献
 
收稿日期 2019/5/13
修回日期 2019/7/8
页码 2600-2603
中图分类号 TP301.6
文献标志码 A