Data Preparation and Structural Models for Web Usage Mining

International audienceThe World Wide Web (or Internet) implies from few years different uses. Each cybernaut wants, on the same medium, to retrieve information, to make shopping, to manage his/her bank accounts, etc. Customers or users want more and more features from the web. In order to realize services with high quality, to evaluate them or simply to design well a website, we want to know how surf or browse a user. Web usage mining is the field that deals with this kind of problems: automatically extract information about users, and then build knowledge about web usage or learn user behavior. In this paper we present a first required step which is the data pre-processing, with a small discussion about different kind of data can be accessible from the internet. Also, we develop our own session retrieving process and we show that structural model as automata can be easily used in a short task of usage prediction. Then, we carry out a set of experiments which show how our pre-processing method can outperform results on artificial data. Keywords: Web usage mining, data pre-processing, grammatical inference