Identifying Glucose Metabolism Status in Nondiabetic Japanese Adults Using Machine Learning Model with Simple Questionnaire

Table 3

Performances of the models for identifying glycometabolic category (95% confidence intervals).

Model	AUC for classifying category 1 and the others	AUC for classifying category 2 and the others	AUC for classifying category 3 and the others	AUC for classifying category 4 and the others	Mean of AUCs	Sensitivity to detect categories 2, 3, and 4	Specificity to detect category 1
Decision tree	0.63 (0.58-0.70)	0.68 (0.60-0.75)	0.56 (0.45-0.66)	0.61 (0.53-0.70)	0.62	0.71	0.41
Support vector machine	0.64 (0.57-0.70)	0.65 (0.57-0.73)	0.58 (0.47-0.68)	0.55 (0.48-0.64)	0.61	0.70	0.55
Random forest	0.69 (0.63-0.74)	0.68 (0.61-0.76)	0.63 (0.55-0.72)	0.67 (0.59-0.74)	0.67	0.70	0.46
XGBoost	0.62 (0.56-0.68)	0.58 (0.50-0.66)	0.59 (0.49-0.69)	0.60 (0.52-0.68)	0.60	0.70	0.45