Fig. 2

Feature importance and model performance for DKD patients classified by eGFR. A Feature importance of the model; B ROC curve and AUC value for the random forest model constructed with full features (n = 23); C ROC curve and AUC value for the random forest model constructed with selected features (n = 6) AUC area under the curve, BMI body mass index, CHD coronary heart disease, DBP diastolic blood pressure, DF diabetic foot, DKD diabetic kidney disease, DR diabetic retinopathy, eGFR estimated glomerular filtration rate, HbA1c hemoglobin A1c, HDL-C high-density lipoprotein cholesterol, LDL-C low-density lipoprotein cholesterol, Lp(a) lipoprotein(a), ROC receiver operating characteristic, SBP systolic blood pressure, SUA serum urea acid, TC total cholesterol