Early antidepressant treatment response prediction in major depression using clinical and TPH2 DNA methylation features based on machine learning approaches

Table 3 Averaged prediction metrics for each classifier based on clinical characteristics

Feature selection	Classifier	ROC (95%CI)	F-Measure	G-Mean	Accuracy	Sensitivity	Specificity	PPV	NPV
Without RFE	Logistic	0.551 (0.529–0.572)	0.710	0.428	0.589	0.813	0.225	0.630	0.426
	Rpart	0.508 (0.496–0.521)	0.684	0.422	0.563	0.766	0.233	0.618	0.380
	SVM-RBF	0.510 (0.498–0.523)	0.762	0.113	0.617	0.990	0.013	0.619	0.445
	LogitBoost	0.569 (0.557–0.580)	0.724	0.523	0.624	0.786	0.348	0.662	0.501
	RF	0.581 (0.569–0.593)	0.687	0.471	0.577	0.750	0.296	0.633	0.422
With RFE	Logistic	0.567 (0.545–0.588)	0.722	0.444	0.604	0.830	0.237	0.638	0.462
	Rpart	0.512 (0.500-0.525)	0.685	0.428	0.565	0.766	0.239	0.620	0.386
	SVM-RBF	0.529 (0.516–0.541)	0.758	0.177	0.615	0.975	0.032	0.620	0.441
	LogitBoost	0.576 (0.564–0.587)	0.726	0.539	0.629	0.782	0.372	0.669	0.513
	RF	0.612 (0.601–0.624)	0.694	0.487	0.587	0.756	0.314	0.641	0.442

ISSN: 1471-244X