一种基于动态时间阈值的会话识别方法  

METHOD OF SESSIONS' IDENTIFICATION BASED ON DYNAMIC THRESHOLD OF TIME

作　　者：戴智丽[1] 王鑫昱[1] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004

出　　处：《计算机应用与软件》2010年第2期244-246,共3页Computer Applications and Software

摘　　要：会话识别是Web日志挖掘的关键步骤,会话识别的质量直接影响后续挖掘的准确性。在Timeout方法固定时间阈值的基础上,提出动态时间阈值,通过对样本日志的分析,得到不同时段的时间阈值。在处理日志文件时,根据当前会话开始记录的访问时间选择时间阈值。实验表明,该方法识别会话的质量比Timeout方法有了明显提高。The sessions＇ identification is a key step in Web log mining. The accuracy of post-mining is influenced by the quality of the sessions＇ identification directly. In this paper, based on fixed threshold of time in Timeout, dynamic threshold of time is proposed. With the sample of log analysed, thresholds of different time are obtained. While log file is processed to identify sessions, the threshold of time is selected according to the access time of the beginning record in current session. The quality of sessions identified by this method is improved obviously than the Timeout method as showed by the experiment.

关 键 词：WEB日志挖掘 数据预处理 会话识别 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置] O213.1[自动化与计算机技术—控制科学与工程]