Evaluating web access log mining algorithms: A cognitive approach

The wide availability of web access logs makes them ideal data sources for the mining of web behavior for electronic commerce competitiveness. Unfortunately, since such logs were originally meant for debugging purposes, they cannot be used directly for mining. Hence, much work has been done to preprocess the logs into a suitable form and subsequently mine them. However such existing techniques make various assumptions that are valid only for specific situations. In addition, there is no fair way to compare them objectively. In this paper we design a framework termed Web Access Log Mining AlGorithm Evaluator (WALMAGE) to represent electronic commerce scenarios and web access log mining algorithms in a way that facilitates the choice of the most appropriate algorithm to use in a particular scenario. We propose a cognitive approach to model our framework so that a wide range-of user behavior can be quantified in order to imbue our framework with realism and practicality.