Use of Machine-Learning Methods to Develop Stress Detection Model

OpenAlex · İngilizce

Human stress detection using physiological signals collected via wearable devices can help patients to reduce the negative impact of stress in the long term. Detecting stress during physiotherapy, especially in remote physiotherapy can guide the healthcare program in rearranging it precisely. This study aims to develop a stress detection model that can be used during remote physiotherapy. The Stress-Predict Dataset, which consists of physiological information (electrodermal activity (EDA), blood volume pulse (BVP), skin temperature (TEMP) from healthy participants, is used in this study to develop the stress detection model. The signal processing, signal filtering, feature extraction and feature selection methods are used to feed the processed signal to machine learning models like the Logistic Regression (LR), Support Vector Machines (SVM), k-nearest neighbours (KNN) and XGBoost. Also Multilayer Perceptron (MLP) deep learning model is used for better performance. Among these algorithms, XGBoost was found to be the best model for the dataset used for the current study, with 98% accuracy.