Table 1:

Performance of classification models on training and reserved test dataset. Boldface signifies that DL method is statistically significantly better in the metric, compared to other methods.

Dataset	Algorithm	Accuracy	SENS	SPEC	F1 Statistic	Balanced accuracy	Computing time/run (sec)
Training	DL	0.909	0.978	0.747	0.952	0.777	570.68
	GBM	0.906	0.600	0.945	0.666	0.772	8.291
	LDA	0.700	0.583	0.718	0.478	0.651	3.118
	LOG	0.906	0.608	0.946	0.681	0.777	5.394
	RF	0.892	0.568	0.946	0.648	0.757	21.340
	RPART	0.801	0.605	0.895	0.620	0.750	3.525
	SVM	0.905	0.663	0.920	0.688	0.791	4.941
Testing	DL	0.912	0.954	0.688	0.930	0.747	1.844
	GBM	0.878	0.560	0.939	0.639	0.749	0.0152
	LDA	0.745	0.627	0.754	0.527	0.691	0.0149
	LOG	0.873	0.550	0.943	0.634	0.747	0.0184
	RF	0.870	0.578	0.938	0.643	0.758	0.0181
	RPART	0.767	0.609	0.861	0.589	0.735	0.0257
	SVM	0.883	0.653	0.927	0.693	0.790	0.0218