This work develops machine learning models for depression categorization using a complete range of characteristics from user-generated data. Physiological data, self- reported symptoms, and demographic data. Predictive models are created using XGBoost, Random Forest, Gradient Boosting, MLPClassifier, and AdaBoostClassifier. Data preparation, feature selection, and model optimization rigorously assess the models. Cross-validation ensures resilience and generalization. Learning curves evaluate models' accuracy, training, and validation loss. Machine learning algorithms properly identify depression levels. Random Forest (99%), Gradient Boosting (93%), MLPClassifier (92%), and AdaBoostClassifier (83%). Learning curves show converging training loss, improving accuracy with more iterations, and constant validation performance. This machine learning study advances mental health categorization. The models help identify and treat depression problems early, enabling tailored care. The research emphasizes feature selection and algorithm choice in model performance. In Conclusion, Machine learning can classify
depression levels. The models improve mental health treatment with accurate and quick depression assessment decision support systems. Expanding the dataset and adding characteristics may enhance model performance and applicability in future studies.