Fig. 4From: Classification of painful or painless diabetic peripheral neuropathy and identification of the most powerful predictors using machine learning models in large cross-sectional cohortsPooled model specific variable importance across resamplings and imputations. Variable importance has been scaled from 0 to 100 to be directly comparable, coloured bars represent the mean, solid black lines the pooled SD calculated using Rubin’s rules, dots represent the variable importance estimated for each imputation. A,C: Mean reduction of classification accuracy when removing each independent variable in Adaptive Regression Splines. B, E: Gini coefficient that quantifies mean decrease of node impurity of the nodes produced by a split that uses each variable in a Random Forest. C: ROC analysis for each independent variable in Naive Bayes classifierBack to article page