Shapley Value Based Variable Interaction Networks for Data Stream Analysis

Jan Zenisek, Sebastian Dorl, Dominik Falkner, Lukas Gaisberger, Stephan M. Winkler, Michael Affenzeller

Abstract: Due to the growing use of machine learning models in many critical domains, ambitions to make the models and their predictions explainable have increased recently significantly as new research interest. In this paper, we present an extension to the machine learning based data mining technique of variable interaction networks, to improve their structural stability, which enables more meaningful analysis. To verify the feasibility of our approach and it’s capability to provide human-interpretable insights, we discuss the results of experiments with a set of challenging benchmark instances, as well as with real-world data from energy network monitoring.