Keystroke Pattern Analysis for Cognitive Fatigue Prediction using Machine Learning

Cognitive fatigue reduces accuracy, productivity, and mental efficiency during prolonged tasks. Traditional detection methods, such as wearables and self-reports, are often intrusive and unreliable. Keystroke dynamics provide a passive alternative by analyzing key press duration, inter-key latency, and typing rhythm. In this work, temporal and statistical features were extracted and classified using Support Vector Machine, Random Forest, and Neural Network models. Preprocessing techniques enhanced data quality and improved model performance. Among the tested approaches, the Neural Network achieved the best results with 91% accuracy, 88% precision, and 87% recall, outperforming Random Forest (89% accuracy) and SVM (87% accuracy). These findings confirm the feasibility of keystroke-based analysis for real-time fatigue detection. The proposed system can be integrated into digital workplaces, e-learning platforms, and healthcare to promote sustained performance and well-being