Cognitive Workload Detection via Binary Chimp Optimization Algorithm and Machine Learning

Detecting cognitive workload during mental tasks is essential for understanding neural activity responses. Electroencephalograms (EEG) are effective tools for this purpose, particularly in mental arithmetic tasks (MAT). This study utilizes EEG data from public databases, focusing on short-term EEG signals to evaluate cognitive workload. The approach employs circulant singular spectrum analysis (Ci-SSA) to decompose EEG signals into intrinsic mode functions (IMFs), followed by entropy-based feature extraction from these IMFs. Feature selection is performed using the binary chimp optimization algorithm (BCOA), and classification is conducted using supervised machine learning algorithms. The proposed method achieves performance metrics, including an accuracy (AR%) of 96.37, sensitivity (SN%) of 97, precision (PR%) of 96, specificity (SE%) of 96, and F1-Score (F1-S%) of 97. The combination of Ci-SSA for signal decomposition and BCOA for feature selection represents a novel approach to cognitive workload detection. The (Ci-SSA+BCOA+KNN) framework demonstrates high classification accuracy for MAT datasets, offering enhanced precision in cognitive workload detection compared to existing techniques.