融合评分相对差异的协同过滤推荐算法  

Collaborative Filtering Algorithm Combined with Relative Differences in Scores

作　　者：王丽萍 傅攀 邱飞岳[2] 陈宏[2] WANG Li-ping;FU Pan;QIU Fei-yue;CHEN Hong(College of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310023,China;College of Education,Zhejiang University of Technology,Hangzhou 310023,China)

机构地区：[1]浙江工业大学计算机科学与技术学院,杭州310023 [2]浙江工业大学教育科学与技术学院,杭州310023

出　　处：《小型微型计算机系统》2022年第7期1388-1393,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61472366,61379077)资助;浙江省科技发展计划重点项目(2018C01080)资助;浙江省基础公益研究计划项目(LGF21F020016)资助.

摘　　要：协同过滤推荐算法是个性化推荐系统中最常用的方法之一,其中相似度计算直接影响基于内存的协同过滤推荐算法的推荐质量.针对协同过滤推荐算法中传统的用户间相似度计算方法仅考虑共同评分项评分数值上的差异导致难以准确衡量非偏好评分场景中用户间相似度的问题,本文提出一种基于余弦相似度并融合评分相对差异的用户间相似度计算方法.该方法考虑评分规模上的差异,计算评分相对相似度并且引入放大系数,在非偏好评分的场景下可以更加准确地区分用户间差异.在真实的数据集上完成对比实验分析,结果表明在非偏好评分场景下,所提方法相较于对比方法能降低预测误差,提高推荐质量.Collaborative filtering recommendation algorithm is one of the most commonly used methods in personalized recommendation systems,in which similarity calculation directly affects the recommendation quality of memory-based collaborative filtering recommendation algorithm.Aiming at the problem that the traditional similarity calculation method between users in the collaborative filtering recommendation algorithm only considers the difference in the scores of the common scoring items,it is difficult to accurately measure the similarity between users in the non-preference scoring scenario.This paper proposes a similarity calculation method between users based on cosine similarity and combined with relative differences in the scores.This method considers the difference in rating scale,calculates the relative similarity of ratings and introduces a magnification factor,which can more accurately distinguish the differences between users in the non-preference scoring scenario.Completing the experimental analysis on the real data set,the result show that the proposed method can reduce the prediction error and improve the recommendation quality compared with the comparison method in the non-preference scoring scenario.

关 键 词：推荐算法 协同过滤 用户相似度 非偏好评分 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]