Ayanika Paul

Machine Learning Intern at Department of Computer Science, MIT Manipal

Engineered and evaluated ML models (XGBoost, Random Forest, Decision Tree, and Logistic Regression) for early detection of diabetes using clinical data; achieved up to 96.8% accuracy and 0.96 ROC-AUC with Random Forest on all features, balancing performance and interpretability. Demonstrated trade-offs between model complexity and explainability; used top 5 features in Random Forest to retain 96% accuracy with a macro F1-score of 0.86, highlighting the feasibility of interpretable ML for healthcare prediction. Also co-authored a research paper on scalable and interpretable diabetes prediction currently under review.