Explainable Hybrid PSO-SVM Framework for EEG-Based Cognitive Confusion Detection

Cognitive confusion is an urgent affective condition that affects learning effectiveness and decision-making performance. Identification of such states with the help of electroencephalography (EEG) helps in human cognitive processes and allow the adaptation of educational or neurofeedback systems. This study proposes an explainable hybrid approach to Particle Swarm Optimization-Support Vector Machine (PSO-SVM) to detect the presence of confusion states in a dataset of publicly available Wanghaohan Confused Student EEG Brainwave Dataset with sustainable accuracy and interpretability. In this study, EEG data recorded through a single-channel NeuroSky MindSet device, which has power spectral features, such as delta, theta, alpha, beta, and gamma, as well as the indices of attention and meditation during the e-learning activities. PSO is used in the proposed framework to optimize the SVM hyperparameters and subsets of features, which guarantees better generalization and less overfitting. The model attained high classification rates in identifying the confounded and non-confounded cognitive states compared to baseline machine learning classifiers. Moreover, Explainable Artificial Intelligence (XAI) models, such as SHapley Additive exPlanations (SHAP) and feature importance analysis, were combined to understand the value of each EEG feature in the model decision. The findings prove that high-frequency gamma and attention scores are the prevailing signs of confusion. The proposed PSO-SVM-XAI system is accurate as well as transparent, providing scalability of intelligent human-computer interaction and adaptive learning systems based on monitoring the cognitive state.