RGIFE: a ranked guided iterative feature elimination heuristic for the identification of biomarkers

Lazzarini, Nicola; Bacardit, Jaume

doi:10.1186/s12859-017-1729-2

BMC Bioinformatics

Table 7 Accuracies and ranks (in brackets) obtained by each method across the ten datasets using four classifiers

From: RGIFE: a ranked guided iterative feature elimination heuristic for the identification of biomarkers

Class.	Method	Prostate-Sbo.	Dlbcl	Leukemia	CNS	Colon-Breast	AML	Prostate-Sin.	Pancreas	Bladder	Breast	Avg. Rank
RF	RGIFE-Union	0.723 (6)	0.643 (3)	0.927 (7)	0.617 (5)	0.947 (3)	0.667 (3)	0.923 (3)	0.898 (3.5)	0.794 (4)	0.884 (2.5)	4.00 (4)
	RGIFE-Max	0.727 (4)	0.573 (7)	0.940 (5.5)	0.600 (6)	0.947 (3)	0.680 (1)	0.913 (6.5)	0.859 (8)	0.782 (6)	0.844 (8)	5.50 (6)
	RGIFE-Min	0.712 (8)	0.680 (1)	0.886 (8)	0.589 (7)	0.927 (7)	0.577 (8)	0.884 (8)	0.900 (1.5)	0.770 (8)	0.851 (7)	6.35 (8)
	CFS	0.741 (1)	0.627 (4)	0.957 (2.5)	0.622 (4)	0.947 (3)	0.597 (7)	0.922 (5)	0.886 (6)	0.800 (3)	0.869 (6)	4.15 (5)
	SVM-RFE	0.733 (2)	0.623 (5)	0.944 (4)	0.668 (3)	0.887 (8)	0.675 (2)	0.923 (3)	0.898 (3.5)	0.819 (1)	0.877 (4)	3.55 (2)
	ReliefF	0.726 (5)	0.577 (6)	0.961 (1)	0.681 (2)	0.930 (6)	0.633 (5)	0.932 (1)	0.900 (1.5)	0.800 (2)	0.892 (1)	3.50 (1)
	Chi-Square	0.716 (7)	0.660 (2)	0.940 (5.5)	0.520 (8)	0.947 (3)	0.622 (6)	0.913 (6.5)	0.886 (6)	0.776 (7)	0.870 (5)	5.60 (7)
	L1	0.730 (3)	0.520 (8)	0.957 (2.5)	0.684 (1)	0.947 (3)	0.650 (4)	0.923 (3)	0.886 (6)	0.788 (5)	0.884 (2.5)	3.80 (3)
SVM	RGIFE-Union	0.709 (4)	0.523 (4.5)	0.917 (6)	0.565 (4)	0.927 (4)	0.633 (4)	0.895 (5)	0.861 (6)	0.757 (5)	0.892 (3)	4.55 (4)
	RGIFE-Max	0.716 (1)	0.573 (2)	0.957 (3.5)	0.572 (3)	0.947 (1.5)	0.617 (5)	0.915 (2)	0.873 (5)	0.775 (2)	0.876 (5)	3.00 (1)
	RGIFE-Min	0.694 (6)	0.567 (3)	0.908 (7)	0.421 (8)	0.907 (5.5)	0.585 (7)	0.852 (8)	0.875 (4)	0.744 (6)	0.894 (1.5)	5.60 (8)
	CFS	0.712 (3)	0.523 (4.5)	0.961 (2)	0.546 (5)	0.943 (3)	0.638 (3)	0.952 (1)	0.911 (1)	0.770 (3.5)	0.832 (8)	3.40 (2)
	SVM-RFE	0.644 (8)	0.500 (6)	0.942 (5)	0.699 (2)	0.850 (7)	0.500 (8)	0.894 (6)	0.848 (8)	0.770 (3.5)	0.894 (1.5)	5.50 (7)
	ReliefF	0.690 (7)	0.407 (8)	0.886 (8)	0.535 (6)	0.747 (8)	0.590 (6)	0.904 (4)	0.900 (2)	0.795 (1)	0.886 (4)	5.40 (6)
	Chi-Square	0.705 (5)	0.603 (1)	0.957 (3.5)	0.462 (7)	0.947 (1.5)	0.652 (1)	0.893 (7)	0.857 (7)	0.739 (8)	0.869 (6)	4.70 (5)
	L1	0.716 (2)	0.490 (7)	0.988 (1)	0.746 (1)	0.907 (5.5)	0.650 (2)	0.913 (3)	0.896 (3)	0.740 (7)	0.851 (7)	3.85 (3)
GNB	RGIFE-Union	0.701 (2)	0.623 (1)	0.932 (5)	0.650 (3)	0.963 (2.5)	0.627 (4)	0.922 (4)	0.887 (5)	0.733 (6)	0.884 (4)	3.65 (1)
	RGIFE-Max	0.698 (3)	0.590 (3)	0.932 (5)	0.620 (5)	0.963 (2.5)	0.610 (6.5)	0.922 (4)	0.846 (8)	0.727 (7)	0.876 (6)	5.00 (7)
	RGIFE-Min	0.691 (5)	0.503 (5)	0.890 (8)	0.589 (7)	0.907 (6.5)	0.617 (5)	0.895 (7)	0.900 (2)	0.751 (4)	0.900 (3)	5.25 (8)
	CFS	0.690 (6)	0.520 (4)	0.973 (1)	0.665 (2)	0.927 (5)	0.650 (2)	0.932 (1.5)	0.871 (7)	0.740 (5)	0.870 (7)	4.05 (3)
	SVM-RFE	0.655 (7)	0.450 (8)	0.958 (3)	0.626 (4)	0.873 (8)	0.593 (8)	0.912 (6)	0.898 (3.5)	0.807 (1)	0.907 (1)	4.95 (6)
	ReliefF	0.616 (8)	0.473 (7)	0.919 (7)	0.570 (8)	0.907 (6.5)	0.633 (3)	0.932 (1.5)	0.898 (3.5)	0.764 (2)	0.901 (2)	4.85 (5)
	Chi-Square	0.694 (4)	0.593 (2)	0.932 (5)	0.620 (6)	0.963 (2.5)	0.610 (6.5)	0.922 (4)	0.925 (1)	0.727 (8)	0.878 (5)	4.40 (4)
	L1	0.719 (1)	0.500 (6)	0.971 (2)	0.746 (1)	0.963 (2.5)	0.655 (1)	0.885 (8)	0.886 (6)	0.753 (3)	0.855 (8)	3.85 (2)
KNN	RGIFE-Union	0.698 (1)	0.593 (2)	0.901 (7.5)	0.665 (4)	0.947 (2.5)	0.602 (3)	0.894 (2.5)	0.911 (1.5)	0.788 (3)	0.876 (6)	3.30 (1)
	RGIFE-Max	0.684 (3)	0.597 (1)	0.927 (3)	0.670 (3)	0.947 (2.5)	0.582 (5)	0.893 (4.5)	0.861 (8)	0.806 (1)	0.862 (8)	3.90 (2)
	RGIFE-Min	0.684 (4)	0.587 (3)	0.901 (7.5)	0.635 (5)	0.947 (2.5)	0.528 (8)	0.903 (1)	0.886 (5.5)	0.758 (8)	0.876 (6)	5.05 (7)
	CFS	0.669 (6)	0.407 (8)	0.917 (5.5)	0.587 (8)	0.910 (6)	0.580 (6)	0.884 (6)	0.911 (1.5)	0.776 (5)	0.876 (6)	5.80 (8)
	SVM-RFE	0.662 (7)	0.523 (7)	0.946 (2)	0.771 (1)	0.847 (8)	0.562 (7)	0.875 (7)	0.898 (3.5)	0.801 (2)	0.892 (1)	4.55 (4.5)
	ReliefF	0.698 (2)	0.553 (5)	0.917 (5.5)	0.615 (7)	0.907 (7)	0.630 (2)	0.894 (2.5)	0.898 (3.5)	0.783 (4)	0.884 (2)	4.05 (3)
	Chi-Square	0.680 (5)	0.537 (6)	0.919 (4)	0.618 (6)	0.947 (2.5)	0.595 (4)	0.893 (4.5)	0.873 (7)	0.770 (6)	0.877 (3)	4.80 (6)
	L1	0.645 (8)	0.570 (4)	0.973 (1)	0.698 (2)	0.927 (5)	0.713 (1)	0.825 (8)	0.886 (5.5)	0.769 (7)	0.876 (4)	4.55 (4.5)

The highest accuracies and ranks are shown in bold. The last column reports the average ranks across the datasets for each method, in brackets are shown the absolute ranks. The accuracies are rounded to the third decimal but the ranks are based on higher precision. RF random forest, KNN K-nearest neighbour and GNB Gaussian Naive Bayes

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com