Tong Zheng and Randy Goebel, University of Alberta
Within the growing literature on web mining, there is a relatively coherent thread of ideas focused on improvements to web navigation. In this paper we focus on the idea of web usage mining, and present a general framework for deploying the mining results and evaluating the performance improvement. The generalized objects created by the application of learning methods are called Navigation Compression Models (NCMs), and we show a method for creating them and using them to make dynamic recommendations. Of note is the observation that no application of any learning method to web data makes sense without first formulating a goal framework against which that method can be evaluated. This simple idea is typically the missing ingredient of many WWW mining techniques. In this paper we present a simulation-based approach to evaluating the effectiveness of Navigation Compression Models non-intrusively by measuring the potential navigation improvement. We evaluate the improvement of user navigation using a quantitative measure called Navigation Improvement (NI), which indicates whether we are actually "improving" the user’s navigation by reducing the number of hyperlinks traversed to find "relevant" pages.