Machine learning models in Heart Rate Variability based mental fatigue prediction: training on heterogeneous data to obtain robust models

Prolonged period of cognitive performance often leads to mental fatigue, a psychobiological state that increases the risk of injury and accidents. Previous studies trained machine learning algorithms on Heart Rate Variability (HRV) data to detect fatigue in order to prevent its consequences. However, the results of these studies can hardly be generalized due to methodological issues including the use of only one type of cognitive task to induce fatigue that potentially leads to task-specific predictions. In this study, we combined the datasets of three experiments each applying different cognitive tasks for fatigue induction to train algorithms that detect fatigue and predict its severity. We also tested different time window lengths and compared algorithms trained on resting and task-related data. We found that classification performance was best when the support vector classifier was trained on task-related HRV calculated for 5-minutes time window (AUC = .843, accuracy = .761). For the prediction of fatigue severity, elastic net regression showed the best performance when trained on 4-min HRV data and self-reported measures (R2 = .224, RMSE = 17.114). These results indicate that both detection and prediction of fatigue based on HRV are effective when machine learning models are trained on heterogeneous, multi-task datasets.