首页> 外文期刊>International Journal of Innovative Computing Information and Control >A SINGULARITY-BASED USER SIMILARITY MEASURE FOR RECOMMENDER SYSTEMS
【24h】

A SINGULARITY-BASED USER SIMILARITY MEASURE FOR RECOMMENDER SYSTEMS

机译:推荐系统基于奇异性的用户相似性度量

获取原文
获取原文并翻译 | 示例
           

摘要

Collaborative filtering is one of the most widely used methods in personalized recommender systems. The most critical part of collaborative filtering is to compute similarities among users using a user-item rating matrix based on which recommendations can be generated. The traditional methods used to calculate user similarity include Pearson correlation coefficient (PCC) and Jaccard Index. However, since PCC defines user similarity as the linear correlation and Jaccard Index defines user similarity as the proportion of common ratings, the accuracy is not ideal if we use these approaches directly. In this paper, we propose a singularity-based similarity measure to resolve this issue. Specifically, we first improve PCC by incorporating the number of common items rated by two users. We then come up with an improved Jaccard method by considering the rating values issued on the items. Finally, we combine the two improved approaches together in two different ways, aiming to further improve recommendation accuracy. Experimental results on two real-world data sets show that our method achieves superior accuracy.
机译:协作过滤是个性化推荐器系统中使用最广泛的方法之一。协作过滤的最关键部分是使用用户项目评分矩阵来计算用户之间的相似度,并以此为基础生成推荐。用于计算用户相似度的传统方法包括Pearson相关系数(PCC)和Jaccard Index。但是,由于PCC将用户相似度定义为线性相关性,而Jaccard Index将用户相似度定义为常见等级的比例,因此,如果直接使用这些方法,则精度并不理想。在本文中,我们提出了一种基于奇点的相似性度量来解决此问题。具体来说,我们首先通过合并两个用户评分的常见项目数来改进PCC。然后,通过考虑在物品上发布的评级值,我们提出了一种改进的Jaccard方法。最后,我们将两种改进的方法以两种不同的方式结合在一起,旨在进一步提高推荐的准确性。在两个实际数据集上的实验结果表明,我们的方法具有较高的准确性。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号