System for a model based analysis of user interaction patterns within web-applications

Web pages could be understood as graphical user interfaces to domain specific data provided by the internet. The specification of HTML5 facilitates the realization of user interfaces based on open standards and thereby the easy integration of code for user-tracking in already existing web pages. This paper describes a system for the detection and analysis of patterns within the user-traces. The core component of the system is a module that provides stochastic models for capturing patterns. In contrast to the simple detection of patterns within a stream of interaction events a model based approach condenses the patterns to states. The system enables an analyst to define restrictions to the state space and facilitates the interpretation of the found states.