Machine Learning Models of Acute Kidney Injury Prediction in Acute Pancreatitis Patients

Table 4

The ranks of feature importance in XGBoost, RF, and LR for predicting AKI.

Rank	RF	XGBoost	LR
1	APACHE II	IAP	APACHE II
2	IAP	PCT	IAP
3	PCT	APACHE II	LPS
4	CRP	TBIL	TBIL
5	TBIL	TG	PCT

Abbreviations: RF: random forest; XGBoost: extreme gradient boosting; LR: logistic regression; APACHE II: Acute Physiology and Chronic Health Evaluation II; IAP: intra-abdominal pressure; PCT: procalcitonin; CRP: c-reactive protein; TBIL: total bilirubin; TG: triglycerides; LPS: serum lipase.