<![CDATA[This paper aims at comparing the suitability of three machine learning models: LightGBM, CatBoost, and Logistic Regression, to lower misdiagnosis rates for psychiatric disorders. Misdiagnosis in mental health may mean improper treatment and, hence, poor outcomes for patients. Our research aims to determine the most accurate predictive model for mental health condition diagnosis that will lead to improved clinical outcomes. We trained and tested these models on an EEG dataset with patient records that have psychiatric diagnoses labeled. For all the models, evaluation and comparison are made using key performance metrics such as Accuracy, Precision, Recall, and F1-Score. Through the use of these methods, it was shown that LightGBM performed better than CatBoost and Logistic Regression, having achieved higher accuracy and F1 scores, indicating more power to make a difference among different psychiatric disorders. These results suggest that machine learning techniques, especially LightGBM, can greatly increase diagnostic accuracy and reduce misdiagnosis in psychiatric contextual systems. Keywords Machine Learning, Psychiatric Disorder, LightGBM, CatBoost, Logistic Regression.]]>
