基于心理测量学的协同过滤相似度.pdfVIP

ISSN1000—12391CN 计算机研究与发展 11—17771TP of Researchand JournalComputer Development 基于心理测量学的协同过滤相似度方法 胡必云 李舟军 王 君 (北京航空航天大学计算机科学与工程学院 北京100191) 10@yahoo．com．on) (byhu82 toCollaborative Psychometrics—BasedSimilarityApproaches Filtering Hu Biyun，LiZhoujun，andWangJun (School Scienceand 100191) ofComputer Engineering，BeihangUniversity。Beijing Abstract collaborativeisa usedin Neighborhood—basedfilteringpopulartechnique a for canbefurtherdividedintouser—basedanditem-based．Ituser’S systems．It predicts preference anunrateditemsbasedonthe betweentwousersoriterns．However，traditional similarity similarity methodsare todata overcomethedrawback，this novel susceptiblesparsity．To paperproposes methodswhichareless todata on psychometrics—basedsimilarity susceptible showthatthe methodsarebetterfor dataset public proposedsimilarity collaborative can therecommendationand accuracy filtering，andthey improve coverage． words collaborative similarity Key neighborhood—basedfiltering；datasparsity；usersimilarity；item 对于未评分项目的偏好．然而，传统的相似度方法易受数据稀疏影响．为了解决这个问题，提出了基于心 滤，它们可以提高推荐准确性和覆盖度(coverage)． 关键词基于邻居协同过滤；数据稀疏；用户相似度；项目相似度 中围法分类号TP391 collaborative usersiS first andthenusedfor Neighborhood-basedfiltering(CF) computed