A Hybrid Machine Learning Framework for Cognitive Load Detection Using Single Lead EEG, CiSSA and Nature-Inspired Feature Selection

Decomposition of EEG data into IMFs was performed using CiSSA. During the second phase, IMFs investigated characteristics that were based on entropy. The characteristics obtained were then used to examine categories by feeding them into nature-inspired algorithms for selection of features (GA, BPSO, PSO, BBA, and BDA) that were trained using machine learning methods (KNN and SVM). Understanding, focus, mental agility, and the capacity to make decisions are all required for an assortment of jobs. On the other hand, the ability of human brains to comprehend and integrate information is restricted. The cognitive "load" refers to the amount of work that these limited capabilities have to do. The dynamics of the brain are significantly altered when cognitive stress is present, as indicated by neurobiology. Changes in cognitive workload patterns are possible. in accordance with characteristics related to stress, anxiety, or mental health, such as breathing, heart rate variability, and skin in addition to blood pressure and conductivity. Outperforming models such as Bat Optimization with K-Nearest Neighbours (BO-KNN), Stationary Wavelet Transform (SWT), KNN, and Support Vector Machine (SVM), the suggested method—which combines Circulant Singular Spectrum Analysis (Ci-SSA), Binary Dragonfly Algorithm (BDA), and K-Nearest Neighbours (KNN)—achieves an astounding accuracy of 99.74%.