Durga Sri Arumugam

Student Mental Health Prediction

• Developed an end-to-end AI/ML pipeline to analyse student academic, lifestyle, and personal factors and predict Mental Health Risk.
• Performed data cleaning, missing-value handling, duplicate checks, data type conversion, outlier detection, exploratory data analysis, and visualization on the raw dataset.
• Built and evaluated GPA regression and Mental Health Risk classification models using feature encoding, train-test splitting, Logistic Regression, Decision Trees, Random Forest, and Gradient Boosting.
• Applied feature importance analysis, ablation study, cross-validation, GridSearchCV hyperparameter tuning, regularization, and decision-threshold analysis to evaluate and improve model reliability; achieved 0.9219 test ROC-AUC with Logistic Regression.
• Developed an LLM-powered prediction explanation system with structured JSON outputs, PII detection guardrails, temperature-based A/B testing, and a Streamlit web application for interactive predictions and explanations.