. 2020 Nov 24;9:3000113. doi: 10.1109/JTEHM.2020.3040236

TABLE 3. Full Feature Set Classification Results.

Evaluation Metric	ECNN			RF			P-Value
Evaluation Metric	Mean±Std. Dev.	95% CI		Mean±Std. Dev.	95% CI		P-Value
TPR	0.746±0.036	0.719	0.773	0.746±0.005	0.742	0.750	0.986
TNR	0.762±0.043	0.729	0.794	0.714±0.007	0.709	0.718	0.004
PPV	0.766±0.024	0.748	0.785	0.710±0.005	0.706	0.713	2.27E-06
NPV	0.744±0.019	0.730	0.758	0.750±0.003	0.747	0.752	0.404
Accuracy	0.755±0.005	0.750	0.757	0.729±0.002	0.728	0.731	2.61E-11

Comparative analysis of predictive performance between ECNN and random forest - a traditional classification model with the capability to perform feature ranking and selection. As shown, ECNN provides statistically significant, superior accuracy whilst providing superior feature selection (see table VII)