基于用户模糊相似度的协同过滤算法user fuzzy similarity-based ....pdfVIP

第37 卷第1 期 通 信 学 报 Vol.37 No.1 2016 年1 月 Journa l on Communications January 2016 doi:10.11959/j .issn.1000-436x.2016024 基于用户模糊相似度的协同过滤算法 吴毅涛，张兴明，王兴茂，李晗 (国家数字交换系统工程技术研究中心，河南 郑州 450002) 摘 要：针对离散评分不能合理表达用户观点和传统协同过滤算法存在稀疏性等问题，借鉴年龄模糊模型，提出 了梯形模糊评分模型。该模型将离散评分模糊化为梯形模糊数，考虑了评分模糊性和信息量，通过梯形模糊数来 计算用户相似度，据此设计了协同过滤算法，并证明了该算法是传统协同过滤算法在模糊域的扩展。实验表明， 该算法在数据稀疏且用户数远多于项目数时性能突出，并且算法运行时间远小于传统协同过滤算法。 关键词：协同过滤；梯形模糊评分模型；模糊距离；模糊相似度 中图分类号：TP393 文献标识码：A User fuzzy similarity-based collaborative filtering recommendation algorithm WU Yi-tao, ZHAN G Xing-ming, WAN G Xing-mao , LI Han （N ationa l D igita l Sw itching System Engineer ing and Techno lo gica l R D Center, Zhengzhou 450002, China) Abstract: In order to reflect the actual case of human decisions and solve the data sp arseness prob lem of traditional col- lab orative filt ering recommendation algorithm, a trap ezo id fuzzy model b ased on age fuzzy model w as proposed . In this model, crisp p oint w as fuzzified into tr ap ezoid fuzzy mb er and th e fuzziness and information of users ’ grade was taken into account when calcu lating user ’s similarity by trap ezo id fuzzy numb er. Based on this model, the user fuzzy similari- ty-b ased collab orative filt ering recomm endation algorithm w as designed . The algorithm was proved to b e an ext ension of traditional collab orativ e filtering algorithm in fuzzy fields. The exp erimental resu lts show that, the proposed algorithm p erforms b etter when imp lem ent ed in the sp arse dataset with more user than item, and its run