The final post in my sepsis prediction series - model results, clinical evaluation, SHAP findings, honest limitations, and what I would do differently.
A technical walkthrough of how I cleaned 790,000 rows of ICU data, engineered clinical features, and trained an XGBoost model to predict sepsis onset
Every year, sepsis kills more Americans than breast cancer, prostate cancer, and opioid overdoses combined. Here’s how machine learning could change that.
