Skip to main content
. 2021 Mar 4;28(6):1235–1241. doi: 10.1093/jamia/ocab003

Table 3.

Model performance based on number of features (14 and 5 features) and method used (NIDDK test dataset)

Performance Logistic Regression CART Random Forest XGBoost
14-Feature Model
AUC 77% 72% 82% 82%
Accuracy 73% 70% 75% 75%
Precision 79% 76% 80% 81%
Sensitivity 79% 78% 82% 81%
5-Feature Model
AUC 75% 73% 78% 79%
Accuracy 73% 69% 70% 74%
Precision 78% 77% 77% 80%
Sensitivity 80% 75% 77% 80%

Abbreviations: Classification and regression trees (CART), area under the curve (AUC).