Mental health issues are a significant worldwide issue that impact people from a variety of socioeconomic and ethnic backgrounds. Effective intervention and treatment planning depend on a timely and precise diagnosis. However, because mental health disorders are inherently complex and variable, traditional diagnostic techniques frequently encounter difficulties. Through the use of sophisticated algorithms that can identify subtle behavioral patterns suggestive of mental illnesses, machine learning (ML) has become a potent tool to support current diagnostic procedures. This study assesses how well several machine learning algorithms-Kernel Naive Bayes, Neural Network, Linear Support Vector Machine (SVM), Logistic Regression, Gaussian Naive Bayes, and Kernel Naive Bayes-classify people into major depressive disorder, mania bipolar disorder, depressive bipolar disorder, PTSD, and normal controls. These models were chosen because of their demonstrated ability to handle a variety of data complexities and produce predictions that are both robust and interpretable. To ensure reliable evaluation, model performances were thoroughly evaluated using key metrics like accuracy, total cost, precision, recall, and F1 score, which were acquired through stratified cross-validation. The findings show that the Neural Network model was the most successful in accurately identifying mental health conditions, with the highest accuracy (89.5%) and balanced precision (0.88) and recall (0.89). These results highlight how machine learning methods can improve diagnostic accuracy and aid in the creation of better clinical mental health assessment instruments and individualized treatment plans.