Comparison of Different Machine Learning Models in Prediction of Postirradiation Recurrence in Prostate Carcinoma Patients

Table 5

Sensitivity, specificity, positive predictive value, negative predictive value and predictive accuracy for ANN and NB models for prediction of disease progression, and comparison between models.

Characteristics	ANN	NB	DeLong’s test
AUC (95% CI)	91.7% (76.2-100%)	97.7% (92.9-100%)	,
Likelihood ratio test^b
ROC cut-off value^c	0.766	0.983	—
Sensitivity (95% CI)	94.4% (72.7-99.9%)	100.0% (80.5-100.0%)
Specificity (95% CI)	91.7% (61.5-99.8%)	92.3% (64.0-99.8%)
PPV (95% CI)	94.4% (72.2-99.1%)	94.4% (72.1-99.1%)
NPV (95% CI)	91.7% (61.9-98.7%)	100%
Predictive accuracy (95% CI)	93.3% (77.9-99.2%)	96.7% (82.8-99.9%)

AUC: area under the curve; CI: confidence interval; NPV: negative predictive value; PPV: positive predictive value; DeLong’s test for two correlated ROC curves.