TABLE 6. Logistic Regression Comparison Using Reduced Feature Selection Results.

Metric	ECNN Features			Full Features			P-Value
	Mean±Std. Dev.	95% CI		Mean±Std. Dev.	95% CI
TPR	0.685±0.008	0.679	0.691	0.286±0.035	0.260	0.313	4.02E-16
TNR	0.711±0.009	0.705	0.718	0.804±0.016	0.792	0.816	0.014
PPV	0.746±0.010	0.738	0.754	0.487±0.032	0.463	0.511	4.63E-14
NPV	0.646±0.007	0.640	0.651	0.633±0.024	0.615	0.651	1.84E-09
Accuracy	0.697±0.005	0.693	0.701	0.600±0.018	0.586	0.613	2.82E-15
AUROC	0.719±0.007	0.717	0.721	0.386±0.034	0.376	0.397	2.68E-61

Shown is a comparative analysis of two logistic regression classification models, one trained on ECNN feature selections and one on the full set of dataset features. As shown, ECNN feature selections result in superior classification performance whilst providing a significant reduction in feature size.