Extracting User Behavior by Web Communities Technology on Global Web Logs

A lot of work has been done on extracting the model of web user behavior. Most of them target server-side logs that cannot track user behavior outside of the server. Recently, a novel way has been developed to collect web browsing histories, using the same method for determining TV audience ratings; i.e., by collecting data from randomly selected users called panels. The logs collected from panels(called panel logs) cover an extremely broad URL-space, and it is difficult to capture the global behaviors of the users. Here we utilize mining results of web community to group those URLs into easily understandable topics. We also use search keywords in search engine sites because user behavior is deeply related to search keyword according to preliminary experiments on panel logs. We develop a prototype system to extract user access patterns from the panel logs and to capture the global behavior based on web communities.