From: Medically-oriented design for explainable AI for stress prediction from physiological measurements
Measurements | Prediction model | Stress prediction accuracy | Paper |
---|---|---|---|
Accelerometer, 34 features from the time and frequency domains of accelerometer data | Naives Bayes, Decision Trees, and Random Forest Classifiers | Highest accuracy 71% using decision trees | [21] |
Accelerometer, GSR, ECG and behavioral features | LDA (Linear Discriminant Analysis)-based classifier | Prediction based on the physiological data and the behavioral features was more accurate than prediction based on physiological data alone | [22] |
Accelerometer, GSR, ECG | Decision Tree Classifier | 92.4% for 10-fold cross validation | [23] |
Accelerometer, video camera, pressure-sensitive touchscreens | J48 tree | 78% in classifying touches as stressed versus not stressed | [24] |
Call logs, Bluetooth data, and SMS data from users’ mobile phones | Random Forest Classifier | 72.39% for binary classification, stressed versus not stressed | [25] |
Physiological data collected from chest-worn and wrist-worn sensors | Deep Convolutional Neural Network | 99.80% accuracy rates for binary classification for stress detection | [27] |