Genome-wide association data classification and SNPs selection using two-stage quality-based Random Forests

BMC Genomics

Table 2 omparison of different random forests models on the SNP pair data sets with different mtry values.

Data set	Model	mtry setting	values	Acc	AUC
ALZ	ts-RF	$\sqrt{M}$	45	.907	.975
	wsRF	log ₂ M	19	.561	.711
	wsRF	(log₂M)²	361	.654	.729
	wsRF	$\sqrt{M}$	616	.692	.757
	GRRF	$\sqrt{M}$	616	.657	.706
	RF	log ₂ M	19	.530	.623
	RF	$\sqrt{M}$	616	.632	.729
	RF	.1M	38,015	.654	.732
	RF	.5M	190,078	.663	.773
	SVM	C	2⁻⁵	.690	.716
PAR	ts-RF	$\sqrt{M_{p}}$	22	.895	.959
	wsRF	log ₂ M	19	.754	.850
	wsRF	$\sqrt{M}$	638	.837	.917
	GRRF	$\sqrt{M}$	638	.688	.765
	RF	log ₂ M	19	.564	.722
	RF	M	368	.799	.848
	RF	.1M	40,880	.808	.879
	RF	.5M	204,402	.827	.898
	SVM	C	2⁻²	.825	.902

ISSN: 1471-2164