Table 1.

Training accuracy of various machine learning algorithms to classify epithelial and mesenchymal cells.

Type	Method	Accuracy (%, $\mu \pm \mathrm{SD}$)
Single	Linear SVM, 1 PC	$93.0\pm 0.2$
	Linear SVM, 2 PCs	$93.0\pm 0.2$
	Linear SVM, 4 PCs	$92.8\pm 0.3$
	Linear SVM, 5 PCs	$93.4\pm 0.6$
	Linear SVM, 6 PCs*	$95.5\pm 0.3$
	Linear SVM, 8 PCs*	$95.1\pm 0.3$
	Linear SVM, 17 PCs*	$95.2\pm 0.6$
	Other SVMs, 6 PCs	$95.4\pm 3.9$
	Decision trees, 6 PCs	$91.6\pm 0.2$
	Nearest neighbor, 6 PCs	$88.9\pm 3.7$
Ensemble	Boosted Trees, 6 PCs, AdaBoost**	$94.3\pm 0.5$
	Bagged Trees, 6 PCs, Bag	$95.2\pm 0.4$

^*$p<0.001$, ANOVA and post-hoc Dunnett’s test versus linear SVM, 1 PC.

^**$p<0.001$, Student’s $t$-test versus linear SVM, 6 PCs.