Problems Modeling Web Sites and User Behavior

Abstract

As the World Wide Web has grown in size and scope, so too has the demand for analysis tools that can help Web site providers determine how their sites are being used. Early analysis approaches focused primarily on accesses to Web documents as recorded in Web server logs. More recent techniques create a model of a site, and the natural modeling approach is to use a directed graph, where pages are denoted by nodes and links are modeled by edges. The process of creating the model and then analyzing the corresponding visitor traffic, however, is fraught with difficulties. The contribution of this paper is a catalog of problems gathered from extensive experience modeling Web sites to determine site structure and analyze user behavior