Frequent sequence mining in web log data

Abstract

The amount of information available even on a single web server can be huge. On the other hand, the amount of visitors (users) can often reach a number of at least six digits. Users vary in gender, age and education, and in consequence their information needs are different. Moreover, they subconsciously expect to get more adequate content after visiting the first few pages. The scope of this kind of problem relates to the domain of information filtering, as a method for delivering relevant information. To solve such a problem, different sources of unstructured or structured data can be used, one of the latter type being web server log data. Executed logging processes on the server side can gather valuable data showing requests sent by users to available resources shared on a particular web site. In this paper, we introduce the Apriori-like FWP algorithm for frequent sequence mining in web log data. Discovered sequences present reconstructed navigation paths across shared web pages by a number of users satisfying a defined minimum. Such knowledge can primarily be used for content recommendation, as well as in cross-marketing strategies and email promotion campaigns.