This project uses EHR and social determinants of health (SDoH) data to predict patient mortality through machine learning, aiming to support early clinical interventions.

Tools & Techniques: Python, Logistic Regression, Random Forest, GridSearchCV, AUC Optimization, SMOTE, Feature Engineering

Patient outcomes vary due to factors like income and access to care. Early risk prediction enables timely, proactive treatment and resource allocation.

We engineered clinical and demographic features to train models predicting mortality. Logistic Regression performed best (AUC = 0.683), and tuning improved Random Forest accuracy by 7.6%.