An interpretable machine learning approach to multimodal stress detection in a simulated office environment

Mara Naegelin, Raphael P. Weibel, Jasmine I. Kerr, Victor R. Schinazi, Roberto La Marca, Florian von Wangenheim, Christoph Hoelscher, Andrea Ferrario

Abstract: Highlights•Our experiment simulates a realistic group office environment and stress scenarios.•Our machine learning models detect stress levels from mouse, keyboard and cardiac data.•Mouse and keyboard data may better detect stress in office contexts than cardiac data.•Specific mouse movement and typing behaviours seem to characterise stress predictions.