Data Science Chair

    Our paper "CompTrails: Comparing Hypotheses across Behavioral Networks" has been accepted at DAMI Journal


    In our paper we propose a method allowing the comparison of human behaviour across behavioural networks with different properties.

    Behavioral Networks is a collective term for networks that contain relational information on human behavior. This ranges from social networks that contain friendships or cooperations between individuals, navigational networks that contain geographical or web navigation, and many more. Understanding the forces driving behavior within these networks can be beneficial in improving the underlying network, for example, by generating new hyperlinks on websites or proposing new connections and friends in social networks. Previous approaches evaluated different hypotheses on these networks and evaluated which hypothesis best fits the network. These hypotheses can represent human intuition, expert opinions, or be based on previous insights. In this work, we extend these approaches to enable the comparison of a single hypothesis across different networks. By understanding the differences in how humans navigate within different networks, it is possible to adjust the structures of the networks and hence create more intuitive and user-friendly experiences. Furthermore, such comparisons can reveal best practices and identify areas for improvement between different networks. We show that naive comparisons do not work and unveil several issues that potentially impact comparisons and lead to undesired results. Based on these findings, we propose a framework with five optional components. Each is applicable in different settings and enables addressing specific analysis goals. We show the benefits of our approach by applying it to synthetic data and several real-world datasets, including web navigation, bibliographic navigation, and geographic navigation.