Web用户聚类算法  

Web Users Clustering Algorithm

作　　者：周宽久[1] 王艳萍[1] 李瑶[1] 

机构地区：[1]大连理工大学系统工程研究所,大连116024

出　　处：《计算机工程与应用》2006年第16期184-186,221,共4页Computer Engineering and Applications

摘　　要：Web用户聚类是将具有相似访问特性的用户归在一起,在电子商务的市场分割和为用户提供个性化服务中,能发挥巨大作用。文章提出基于用户访问路径以及节点高度的相似性评价函数,建立相似矩阵,并提出相似矩阵结构分解算法对此相似矩阵进行变换生成对角矩阵或下三角矩阵,该矩阵中的每一个子对角阵即对应一个用户类,矩阵的左下角对应各用户类之间的关系。Web users clustering is to divide Web users with different access attributes into different groups and all web users in one group with the same access attributes.Clustering is playing an important role in dividing market and providing personal services in E-business.Similarity evaluating function based on access paths and hight of nodes in web structure tree is issued in this paper.A similarity matrix is created according to the valuating function.Next,a matrix decomposition algorithm is put forward to convert the matrix into a diagonal matrix or a lower triangle matrix.A sub-diagonal matrix in the similarity matrix is corresponding to a group of users and the lower triangle matrix means the relationship among the different groups.

关 键 词：WEB用户聚类 相似矩阵 结构分解 模糊聚类 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]