EEG-based detection of cognitive load using VMD and LightGBM classifier

Cognitive load, which alters neuronal activity, is essential to understanding how the brain reacts to stress. This work aims to classify electroencephalogram (EEG) signals to detect cognitive load by extracting features from intrinsic mode functions (IMFs). The variational mode decomposition (VMD) was used for the eight-level decomposition of each EEG channel data (4 s). Next, entropy-based features were extracted from each IMF. The extracted features were fed to supervised machine learning (ML) classifiers: light gradient boosting machine (LightGBM), extreme gradient boosting (XGBoost), and categorical boosting (CatBoost) for classification. Experiments are conducted on two public EEG datasets, multi-arithmetic tasks (MAT) and simultaneous task EEG workload (STEW). The performance is measured via accuracy, specificity, sensitivity, positive predictive value, log-loss score, F1 score, and area under receiver operating curves (AUROC). The proposed LightGBM classifier technique demonstrates superior classification accuracy rates of 97.22% and 95.51% for the MAT and STEW datasets. The experiment results demonstrated that the proposed technique detects cognitive load more precisely than existing methods. The LightGBM classifier model enhanced accuracy and sensitivity in predicting outcomes through the utilization of ML and data mining methods.