Pre-Processing and Clustering Complex Data in E-Commerce Domain

This paper presents our preprocessing and clustering method on a clickstream dataset issued from e-commerce domain. The main contributions of this article are double. First, after presenting the clickstream dataset, we show how we build a rich data warehouse based an advanced preprocessing method. We take into account the intersite aspects in the given e-commerce domain, which offers an interesting data structuration. A preliminary statistical analysis based on such complex data i.e. time period clickstreams is given, emphasing the importance of intersite user visits in such a context. Secondly, we describe our crossed-clustering method which is applied on data generated from our data warehouse. Our preliminary results are interesting and promising illustrating the benefits of our WUM methods, even if more investigations are needed on the same dataset.