Sequence Mining on Web Access Logs: a Case Study

We present a case study in which sequence mining algorithms were applied to web access log data. The data are from a portal that is targeted for business users. In this portal, like in many others, content is described using a set of descriptors, such as keywords, category and type. We investigate whether representing content by the type rather than its identifier enables existing sequence mining methods to obtain interesting patterns. Rather than a more traditional approach based on measures such as support and confidence, we analyze results from an application perspective. This enables us to identify opportunities for improving and extending these methods.