A Novel Integrating Multimodel Sensor Data with Machine Learning for Stress Detection

Stress is a common aspect of modern life that affects bothphysical and mental well-being. While too much stress can have negativeconsequences, manageable levels of stress can actually foster creativityand problem-solving skills. Therefore, it’s important to monitor and man-age stress in order to maintain good health, which highlights the needfor accessible mental health support programs. Wearable devices likesmartwatches offer promising solutions by measuring physiological indi-cators such as electrodermal activity (EDA) to assess stress levels. Thisstudy aimed to investigate using indicators from wearable devices wornon the wrist to predict stress levels. We analyzed data from the WESADdataset and used seven machine learning algorithms, including Logis-tic Regression (LR), Random Forest (RF), K-Nearest Neighbors (KNN),and Decision Tree (DT). It is interesting to note that both Random For-est (RF) and Decision Tree (DT) algorithms achieved an accuracy rateof 99% for both sets of features - All Features and Extracted Features.