Classifying Cognitive Workload Using Machine Learning Techniques and Non-Intrusive Wearable Devices

Effective management of cognitive workload is essential to ensure user performance and minimize mental fatigue. This study aimed to evaluate the effectiveness of three machine learning (ML) algorithms, including a Support Vector Machine (SVM), a Random Forest (RF) and Random Fourier Features (RFF) models, for predicting instances of high cognitive workload based on physiological data collected from non-intrusive wearable devices. The study included scenarios to induce high workload. After processing data from 30 human participants, and extracting cognitive workload-related features, we evaluated the ML algorithms in terms of accuracy, Area Under the Curve (AUC), and Average Precision (AP) responses. The RFF model emerged as a top performer, achieving 97% accuracy, 99.5% AUC, and 99% AP, demonstrating the capability to detect high workload scenarios and robustness to unbalanced data in workload predictions. The most significant physiological features in the RFF model included skin conductance level (SCL) change, SCL mean, and percent change in pupil size. Future studies should enhance the present binary classification model to support multi-class or continuous workload predictions and implement real-time analysis using trained algorithms.