Explainable machine learning model for predicting the risk of significant liver fibrosis in patients with diabetic retinopathy

Table 3 Predictive performance of eight machine learning models on the validation set

Model	AUC	Optimal cutoff	Accuracy	Sensitivity	Specificity	Precision	F1 score	Brier score
LR	0.850 (0.829, 0.869)	0.409	0.779 (0.758, 0.799)	0.770 (0.734, 0.803)	0.785 (0.759, 0.810)	0.704 (0.671, 0.740)	0.736 (0.709, 0.763)	0.153 (0.143, 0.164)
RF	0.837 (0.817, 0.857)	0.478	0.790 (0.771, 0.810)	0.721 (0.686, 0.754)	0.836 (0.812, 0.859)	0.746 (0.712, 0.780)	0.732 (0.704, 0.761)	0.158 (0.148, 0.169)
NB	0.834 (0.813, 0.854)	0.209	0.770 (0.749, 0.791)	0.721 (0.687, 0.755)	0.802 (0.776, 0.827)	0.708 (0.673, 0.743)	0.714 (0.688, 0.742)	0.158 (0.148, 0.169)
SVM	0.834 (0.813, 0.856)	0.388	0.770 (0.750, 0.790)	0.750 (0.715, 0.783)	0.784 (0.758, 0.810)	0.698 (0.663, 0.730)	0.723 (0.695, 0.749)	0.159 (0.149, 0.171)
XGBoost	0.834 (0.814, 0.855)	0.387	0.775 (0.754, 0.794)	0.761 (0.728, 0.792)	0.784 (0.758, 0.809)	0.702 (0.666, 0.736)	0.730 (0.704, 0.756)	0.161 (0.149, 0.173)
MLP	0.807 (0.784, 0.828)	0.448	0.752 (0.730, 0.773)	0.720 (0.685, 0.754)	0.774 (0.746, 0.800)	0.680 (0.642, 0.716)	0.699 (0.669, 0.725)	0.186 (0.180, 0.192)
KNN	0.779 (0.757, 0.802)	0.265	0.697 (0.672, 0.720)	0.789 (0.755, 0.821)	0.635 (0.603, 0.668)	0.590 (0.553, 0.625)	0.675 (0.644, 0.703)	0.205 (0.192, 0.219)
DT	0.749 (0.725, 0.774)	0.473	0.774 (0.754, 0.794)	0.670 (0.635, 0.709)	0.843 (0.819, 0.866)	0.740 (0.706, 0.774)	0.703 (0.675, 0.733)	0.175 (0.163, 0.186)

Abbreviations: XGBoost: Extreme Gradient Boosting; RF: Random Forest; MLP: Multilayer Perceptron; SVM: Support Vector Machine; LR: Logistic Regression; NB: Naive Bayes; DT: Decision Tree; KNN: K-Nearest Neighbors; AUC: Area Under the Curve

ISSN: 1472-6947