Klasifikasi Gangguan Mental Menggunakan Metode TabularBERT Dengan Explainable AI Untuk Model Interpretabilitas

Mental disorders are health issues that require early detection and appropriate intervention. This study develops a mental disorder classification model using the TabularBERT method and evaluates the impact of Principal Component Analysis (PCA) on improving its performance. A dataset of mental disorder symptoms was utilized with a train-test split scheme, and the evaluation employed metrics including accuracy, precision, recall, F1-score, and a confusion matrix visualization. Model interpretability was analyzed using Local Interpretable Model-Agnostic Explanations (LIME). The results indicate that applying PCA consistently enhances model performance. LIME analysis reveals differences in feature contribution between models with and without PCA. This study demonstrates that the combination of TabularBERT, PCA, and LIME not only produces a high-performing classification model but also supports interpretability, making it potentially applicable to decision support systems in the mental health domain to improve the quality of detection and treatment of mental disorders.