Multi-view ensemble learning for EEG-based detection of Obsessive-Compulsive Disorder.

Psychiatric disorders pose a critical challenge in modern healthcare due to their high prevalence, complex symptomatology, and reliance on subjective diagnostic procedures. Obsessive-Compulsive Disorder (OCD), in particular, is a chronic and debilitating condition that significantly impairs daily functioning, yet remains difficult to diagnose accurately and objectively. To address this challenge, we propose a novel frequency-aware multi-view ensemble learning framework for EEG-based OCD classification. The method systematically partitions EEG features into frequency-specific views (delta, theta, alpha, beta, and gamma), thereby capturing complementary neural dynamics associated with different cognitive and affective processes. For each frequency view, multiple base classifiers are trained, and their probabilistic outputs are integrated through a Particle Swarm Optimization (PSO)-based weight optimization strategy. This ensures that more informative frequency bands and classifiers contribute proportionally to the final prediction. Comprehensive experiments conducted on clinical EEG datasets demonstrate that the proposed model achieves superior performance, with an average accuracy of 90.1%, precision of 90.6%, recall of 90.1%, and F1-score of 89.7%, significantly outperforming state-of-the-art baselines such as SVM (70.2% accuracy) and Random Forest (68.7% accuracy). The results establish the clinical potential of this framework as a robust, non-invasive, and scalable computational tool to complement traditional psychiatric assessment and improve diagnostic precision in OCD recognition.