Analysing physiology of interpersonal conflicts using a wrist device

Abstract

We present a study in which 59 participants logged their interpersonal conflicts while wearing an Empatica E4 wristband. They marked the beginnings and endings of the conflicts, as well as their intensity. In this paper, the dataset is described and a preliminary analysis is performed. We describe data segmentation and feature calculation process. Next, the interrelationships between the features and labels are explored. A logistic regression model for conflict recognition was built and significant features were selected. Finally, we constructed a machine learning model and proposed how to improve it.