<![CDATA[Schizophrenia is a mental disorder that affects about 0.3% of the world population. It is characterized by a wide range of symptoms that form several subtypes. Overlapping symptoms and subjective clinical assessments may reduce consistency and make subtype classification challenging. Machine learning algorithms that use patients’ medical records offer a potentially objective approach for subtype classification. This study aims to classify four schizophrenia subtypes: paranoid, catatonic, undifferentiated, and residual, based on subtype labels recorded in the hospital using a multiclass SVM approach with kernel optimization. The dataset consists of 218 medical records of schizophrenia patients with 25 binary symptom variables used as input features. SVM was trained using two multiclass approaches, namely OAO and OAA. Evaluation was performed using five-fold stratified cross-validation. Performance was calculated using accuracy, macro-precision, macro-recall, and macro F1-score. Optimal performance was achieved using the OAA approach with an RBF kernel at C = 10 and gamma = 0.1. This configuration achieved an accuracy, macro-precision, macro-recall, and macro F1-score of 0.89, 0.90, 0.86, and 0.87, respectively. These results show that the multiclass approach, kernel functions, and parameter configuration influence classification performance. The proposed model may serve as a screening or decision-support tool to assist subtype identification based on clinical symptom records.  ]]>
Copyrights © 2026
