Table 2.

Model Performance for Prediction of Overall Glaucoma Surgical Failure

Model	AUROC (95% CI)	Accuracy (95% CI)	F1 (95% CI)	Sensitivity (Recall) (95% CI)	Specificity (95% CI)	PPV (Precision) (95% CI)	NPV (95% CI)
Random Forest	0.767	0.755	0.850	0.955	0.223	0.765	0.660
	(0.730–0.804)	(0.724–0.786)	(0.828–0.870)	(0.940–0.972)	(0.174–0.284)	(0.733–0.796)	(0.552–0.770)
SVM	0.762	0.710	0.794	0.769	0.556	0.820	0.478
	(0.725–0.802)	(0.676–0.745)	(0.764–0.823)	(0.728–0.812)	(0.489–0.620)	(0.788–0.852)	(0.415–0.546)
Gradient Boosting	0.756	0.733	0.820	0.842	0.444	0.800	0.518
	(0.717–0.792)	(0.701–0.765)	(0.795–0.845)	(0.809–0.878)	(0.379–0.507)	(0.768–0.831)	(0.449–0.595)
Gaussian Naïve Bayes	0.674	0.565	0.258	0.152	0.926	0.884	0.293
	(0.631–0.719)	(0.530–0.601)	(0.203–0.306)	(0.116–0.186)	(0.894–0.963)	(0.775–0.912)	(0.259–0.329)
LDA	0.754	0.737	0.827	0.868	0.393	0.790	0.530
	(0.715–0.793)	(0.705–0.767)	(0.802–0.850)	(0. 837–0.898)	(0.321–0.459)	(0.757–0.821)	(0.450–0.609)
Logistic Regression	0.765	0.747	0.846	0.956	0.193	0.757	0.634
	(0.727–0.803)	(0.717–0.776)	(0.824–0.866)	(0.943–0.973)	(0.141–0.242)	(0.725–0.788)	(0.525–0.750)
KNN	0.675	0.735	0.841	0.966	0.126	0.744	0.586
	(0.631–0.719)	(0.704–0.765)	(0.819–0.862)	(0.951–0.983)	(0.084–0.165)	(0.712–0.775)	(0.438–0.762)
Multi-Layer Perceptron	0.734	0.729	0.819	0.848	0.415	0.792	0.509
	(0.696–0.772)	(0.695–0.761)	(0.792–0.844)	(0.813–0.883)	(0.341–0.480)	(0.758–0.824)	(0.435–0.591)
Decision Tree	0.661	0.702	0.820	0.935	0.089	0.730	0.343
	(0.619–0.705)	(0.669–0.735)	(0.795–0.843)	(0.913 – 0.959)	(0.050–0.125)	(0.697–0.762)	(0.217–0.500)
Neural Network	0.766	0.755	0.837	0.870	0.452	0.807	0.570
	(0.735–0.801)	(0.720–0.789)	(0.814 – 0.862)	(0.830–0.913)	(0.462–0.703)	(0.797–0.817)	(0.462–0.667)

CI, confidence Interval; KNN, K-Nearest Neighbors; LDA, linear discriminant analysis; NPV, negative predictive value; PPV, positive predictive value; SVM, support vector machine.