Stress Prediction Based on Chaos Theory and an Event–Behavior–Stress Triangle Model

Ningyun Li,Ling Feng,Huijun Zhang,Lei Cao,Xin Wang
DOI: https://doi.org/10.1109/tcss.2024.3386752
2024-01-01
Abstract:Predicting stress can help people take timely action to manage stress before potential physical and psychological problems arise. In this study, we analyze and verify chaotic features of human's stress response to stressor and uplift events, and present an event-behavior-stress triangle model for stress prediction. We reconstruct the phase space based on chaos theory, and integrate stress-correlated pre and postfactors (events and behaviors) through an event-behavior-stress correlation memory and a behavior-stress correlation memory for stress prediction. User's personal features (including self-cognition, opinion about school, personality traits, and future event's impact) are also involved in stress prediction. We conduct the experiments on the publicly available StudentLife dataset collected from a mobile phone app, including users' daily activities inferred through the automatic and continuous sensing application and users' self-reported ecological momentary assessments (EMA) data. The experimental results show that the proposed method outperforms four baseline methods, achieving (88.13% accuracy, 79.38% precision, 77.10% recall, 78.19% F1-score) for 2-labeled (nonstressed/stressed) stress prediction, and (70.42% accuracy, 69.21% precision, 67.90% recall, 68.53% F1-score) for 3-label (nonstressed/little-stressed/huge-stressed) stress prediction. Further possible improvements and implications related to chaos-based stress prediction are also discussed at the end of the article.
What problem does this paper attempt to address?