On Modelling and Synthetically Generating Web Usage Data

While the whole public Web is a potential source for Web content and Web structure mining, the actual usage information, that is essential for Web usage mining (WUM), is kept hidden by Web servers of hosted Web sites. Furthermore, there are only a handful of poorly described Web access datasets publicly available. On the one hand, the lack of public datasets hamper WUM research, while on the other hand, online services demand for advanced techniques, e.g. to profile their customers and personalize their Web based services. In this paper we propose our methodology to build synthetic Web usage data generators based on the knowledge established by an extensive analysis of five real-world Web usage datasets.