基于数据挖掘技术的网站用户分析

采用数据挖掘中的聚类技术,对微软网站日志文件中的用户行为数据进行分析,在用户行为的基础上将用户归为同质的组,从而寻求一种识别典型访问情况的方法.采用了一种“将SPR-SQ减小的情况屏蔽,只考虑SPRSQ增加的情况”的处理方法,来实现最佳聚类个数K的选择.同时,在计算组内偏差的时候,提出了“冗余组内偏差”的概念.在聚类分析阶段完成之后,对每个聚类结果进行“标准化均值”比较,并对其用户行为作了简要分析.

Analysis of Website User Based on Data Mining

By adopting the Clustering technology in Data Mining,the data of users' behavior within a log file of www.microsoft.com are analysed in this paper,and all the users are classified to several consubstantial groups on the basis of users' behavior,then explored a way to identify typical visiting circumstances.A process method that only the ascendable circumstances of SPRSQ are considered and the delinable circumstances of SPRSQ are shielded, is used to calculate the best K value.Moreover,when having calculated the warp within groups,a concept called "Redundant Warp Within Groups" is put forward.After the stage of Clustering analysis,every Cluster is compared with standardized average value,and the users' behavior of every cluster is then analysed in brief.