论文部分内容阅读
针对网络日志挖掘中的会话切分问题,提出了一种基于时间间隔的方法。该方法在相邻页面访问时间间隔超出某阈值时切分会话,针对特定IP的阈值根据其频率矢量来定义。实验表明:代理服务器IP和单用户IP的频率矢量具有不同特性,代理服务器IP的频率矢量具有Power-law的特点,而单用户IP的频率矢量具有Gauss分布的特点,在此基础上提出一种基于Gauss假设的方法来设定不同单用户IP的阈值。与传统的对所有IP地址使用单一的先验阈值进行切分的方法相比,该方法更为合理有效。
Aiming at the problem of session segmentation in web log mining, a method based on time interval is proposed. This method cuts the conversation when the adjacent page access time interval exceeds a certain threshold, and the threshold for a specific IP is defined according to its frequency vector. The experiments show that the frequency vectors of proxy server IP and single-user IP have different characteristics. The frequency vector of proxy server IP has the characteristics of Power-law. The single-user IP frequency vector has the characteristics of Gauss distribution. Based on this, Gauss hypothesis-based approach to set the threshold for different single-user IP. This method is more reasonable and effective than the traditional method of using a single a priori threshold for all IP addresses.