Multiclass Classification of Neurological and Psychiatric Conditions Using Synthetic Neuroinformatics Biomarkers and EEG Band Simulation

Abstract

This study presents a machine learning framework for multiclass classification of neurological and psychiatric disorders using synthetic neuroinformatics biomarkers and EEG spectral simulations. Synthetic data modeled cognitive-motor features (visual delay, motor gain, expectancy weight, memory capacity, EEG bands), while real EEG data from 50 PhysioNet recordings validated performance. An XGBoost model optimized through grid search (24 combinations, five-fold validation) achieved 97.8% accuracy and 0.978 F1-score. Results showed excellent classification of neurotypical, ADHD, ASD, dementia, depression, GAD, Parkinson’s, psychosis, and Tourette’s, demonstrating strong potential for neuropsychiatric diagnosis support.