A XGBOOST Based Algorithm for Early Prediction of Human Sepsis

Sepsis is a potentially lethal disease that occurs when the body's immune system reacts to an infection by attacking its own tissues. Organs work poorly and incorrectly when the body's infection-fighting mechanisms kick in. Sepsis can lead to septic shock. This is an abrupt drop in blood pressure that can damage organs and possibly cause death. Antibiotics and intravenous fluids should be administered as soon as possible to improve the chances of survival. The aim is to create and validate an XGBoost algorithm for early sepsis prediction and the results are compared with Decision tree, Gradient Boosting Tree(GBT),Random Forest. XGBoost algorithm produces accuracy of 95.01% which outperforms the other three algorithm.