Figure 4
From: Predicting outcomes of acute kidney injury in critically ill patients using machine learning
Feature importance for the top 20 features for CKD prediction after prediction. For each classifier, the feature importance estimation was based on the mean decrease in impurity (MDI) calculations. In the features set, first (abs. change) SCr and second (abs. change) SCr show the absolute change in SCr at the ICU admission from baseline SCr and absolute change in SCr at the AKI event from baseline SCr, respectively.
