Impact of Bio-inspired Optimization Techniques on EEG Signal Classification for Psychiatric Disorder Diagnosis

This study investigates the classification of psychiatric disorders using electroencephalogram (EEG) data, with a focus on enhancing feature selection and classification accuracy through advanced optimization techniques. The methodology comprises several stages: EEG signal acquisition, preprocessing, feature extraction, and clinical-based class labelling. Given the inherent challenges of high-dimensional data and class imbalance in EEG datasets, five bio-inspired optimization algorithms are employed to optimize feature selection: Artificial Bee Colony (ABC), Particle Swarm Optimization (PSO), Monarch Butterfly Optimization (MBO), Harris Hawks Optimization (HHO), and Glowworm Swarm Optimization (GSO). These algorithms aim to identify the most relevant features while minimizing computational overhead. An Artificial Neural Network (ANN) classifier is trained on the optimized feature sets to predict psychiatric disorders. A comparative analysis of the optimization methods reveals that HHO offers the fastest execution time, MBO excels in selecting a diverse set of features, and ABC provides a strong balance between efficiency and classification accuracy. The findings underscore the significance of selecting suitable optimization strategies tailored to computational resources and data characteristics to improve EEG-based psychiatric disorder classification.