Web日志挖掘中的会话识别算法  

Improved method for session identification in web log mining

作　　者：蔡浩[1] 贾宇波[1] 黄成伟[1] 黄志强[1] 

机构地区：[1]浙江理工大学信息电子学院,浙江杭州310018

出　　处：《计算机工程与设计》2009年第6期1321-1323,1390,共4页Computer Engineering and Design

基　　金：浙江省教育厅科研计划基金项目(20060599);浙江理工大学科学基金项目(111251A4Y04002)

摘　　要：会话识别是Web日志挖掘的关键步骤,然而很多方法所得到的会话不够精确。针对Web日志挖掘中的会话识别问题,在最常用的Timeout方法的基础上,提出了一种改进的基于平均时间阈值的识别方法。通过动态计算会话中请求记录间的平均时间间隔,个性化地调整页面的时间阈值,相对于传统的对所有用户页面使用单一的先验阈值,该方法能够更准确地识别出长对话。最后对生成的侯选会话集进行二次识别,使识别出的会话更为合理有效。实验结果表明,会话质量得到了提高。Although session identification is an important step in web log mining, the session identified by existing methods are not accurate. Toward session identification in web log mining, an access timeout-based improvement is carried out of session identification in web log mining. By calculating the average intervals dynamically among request records in the session, adjusting a threshold individually. Compared to the traditional method that defines a uniform threshold for all web pages experimentally, the approach presents can identify the long session more accurately. Then generating sets of candidate session is re-identified, which make the session more reasonable and effective. The quality of session identification is proved more efficiency by experiments.

关 键 词：WEB挖掘 数据预处理 会话识别 时间阈值 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TP393[自动化与计算机技术—计算机科学与技术]