Figure 3.

The ROC comparison of four algorithms based on nine variables. The classifier with GBDT obtains an AUC of 0.975 [95% confidence interval (CI): 0.986–0.963], and has the best performance when comparing with the other three algorithms. ROC, Receiver Operating Characteristic; GBDT, Gradient Boosting Decision Tree; RF, Random Forest; DNN, Deep Neural Networks. Nine items are hemoglobin, serum creatinine, serum calcium, immunoglobulin (A, G and M), albumin, total protein, and ratio of albumin to globulin.