基于树型网络结构的用户相似性度量算法的研究.pdf

l蠹基于树型网络结构的用户相似性度量算法研究信息管理与电子商务专业研究生李雪琴指导教师李聪摘要随着互联网技术的迅速发展,人们可以从互联上获取更多的信息和服务。这些信息在为人们带来方便的同时也给用户带来了择优困难。随着电子商务不断发展,产品数量迅速增长,用户需耗费大量的时间寻找自己想要的产品。在电子商领域,推荐系统不但能提供个性化推荐,还能通过和用户之间建立联系让用户产生依赖。本文首先分别介绍了推荐系统和树型网络结构的研究意义。对现有的树型网络结构相似性度量方法进行分析,将相似性计算方法分为基于分解策略、基于路径比较、基于节点匹配、基于双边匹配法和最大公共子树法的树型网络结构相似性比较方法。通过研究发现,传统的算法存在跨层匹配、没有考虑节点的层次权重等问题。随后根据用户行为数据建立用户树型网络结构并分析其特征,在传统的基于路径匹配树型网络结构相似性算法的基础上提出了基于权重路径分解的用户兴趣树相似性匹配算法。并将该算法应用于传统的协同过滤推荐方法中,修改了用户相似性比较方法。最后通过实验验证,本文提出的度量模型要优于传统方法,达到了预期的效果。关键词:树型网络结构:相似性度量:推荐系统;协同过滤:用户兴趣树万方数据 User similaritymeasuring algorithm based on tree—work Major:informationmanagement and merce Graduate Student:Xueqin L1 Adviser:Cong Li Abstract With thefastdevelopment ofIntemet technology,people can get more information and services from easier forpeople tolive agood lifein such ainformation italso brings difficulties to US the scale of merce isexpanding and thefast-growing number and variety ofgoods,customers need tospend alotoftime tofmd theproducts thattheywant order tosolve those problems,mendation system came into order to solve these problems, personalized mendation system came the area merce,a good mender system not only provides users with a personalized service,but also establish close relationship with are ing rely on mendation system. This paper firstly introduced the Significance of the research on mendation systems and similarity measures oftree· we classified thesimilarity measures of tree into six categories, including operating-strategy-based, position·s打ategy—based,parison-based and parison-based methods, bilateral matching method and largest public subtree traditional matching algorithms have theproblems such ascross-layer matching,not considering thehierarchical weights and SO ,user intereststree—work isbuiltbased on the user behavior and weighted position Tree similaritymatching algorithm isproposed based on thetradit