Abstract: Concerning the problem of data collecting in multi users and multi Web sites access environment, a new collecting and cleaning method is proposed. Network sniffer is used to record user browsing behaviors; combined with the traditional cleaning method, a new filtering method which is based on the referer relations of requests and the intervals of requests is designed, in order to remove the huge number of concomitant requests which generate concomitantly with user activity. Experimental results show that the technique can eliminate concomitant requests and extract base requests which explicitly occur by the user efficiently. These can provide accurate data source for building the user profile.

Key words: Hypertext Transport Protocol（HTTP） request, browsing behaviors, data collecting, data cleaning, filtering

摘要： 针对在多用户、多Web站点的网络访问环境下存在的数据采集问题，提出了一种新的数据采集及清理方法。引入网络嗅探的方式进行浏览行为数据的采集；在结合传统数据清理方法的基础上，提出一种利用HTTP请求间存在的引用和时间关系来清除HTTP请求产生的大量的非用户显式点击的附带请求。实验结果表明，该方法可以有效地采集浏览行为数据并清除大量的附带请求，抽取出能够反映用户显式点击的页面基请求，为用户浏览行为建模提供准确的数据源。

关键词: 超文本传送协议（HTTP）请求, 浏览行为, 数据采集, 数据清理, 过滤