Random KNN feature selection - a fast and stable alternative to Random Forests

BMC Bioinformatics

Table 5 Comparative performance with gene selection, Group I

Dataset	p* c/n	Mean Accuracy			Standard Deviation			Coefficient of Variation
		RF	R1NN	R3NN	RF	R1NN	R3NN	RF	R1NN	R3NN
Ramaswamy	1267	0.577	0.726	0.704	0.019	0.013	0.013	3.231	1.775	1.796
Staunton	859	0.561	0.692	0.663	0.042	0.026	0.031	7.485	3.802	4.669
Nutt	829	0.671	0.903	0.834	0.051	0.030	0.031	7.619	3.268	3.674
Su	792	0.862	0.901	0.888	0.016	0.015	0.014	1.884	1.624	1.567
NCI	688	0.813	0.854	0.836	0.033	0.027	0.023	4.083	3.135	2.796
Brain	666	0.969	0.958	0.940	0.025	0.013	0.018	2.574	1.323	1.875
Armstrong	468	0.936	0.993	0.980	0.020	0.009	0.013	2.166	0.938	1.345
Pomeroy	329	0.858	0.933	0.863	0.025	0.016	0.017	2.892	1.762	1.991
Bhattacharjee	310	0.934	0.956	0.954	0.015	0.006	0.006	1.572	0.620	0.618
Adenocarcinoma	260	0.942	0.939	0.859	0.018	0.017	0.032	1.948	1.808	3.675
Golub	222	0.943	0.986	0.986	0.022	0.003	0.004	2.328	0.289	0.369
Singh	206	0.889	0.952	0.931	0.024	0.014	0.018	2.718	1.427	1.920
Average		0.830	0.899	0.870	0.026	0.016	0.018	3.375	1.814	2.191

ISSN: 1471-2105