Correction to: Geochemical Prospectivity Mapping Through a Feature Extraction–Selection Classification Scheme

Correction to: Natural Resources Research https://doi.org/10.1007/s11053-018-9422-3

The original version of this article unfortunately contained a mistake in the second part of Table 3. The data in last four rows of Table 3, i.e., columns under “Random forests” section have been shifted inadvertently to the previous columns due to a production error. The corrected section of the table is given below.

Machine learning algorithm	Feature extraction (de-noised)	Feature selection	Retained features	Training data (tenfold cross-validation)				Testing data
				Accuracy			ROC	Accuracy			ROC
				AN	BG	Overall	Area	AN	BG	Overall	Area
Random forests	No	None	All 21 original variables	76.1	69.8	73.1	0.79	71.0	77.8	72.4	0.78
	Yes	None	All 21 original variables	78.9	68.3	73.9	0.8	71.0	81.5	75.9	0.78
	Yes	Principal component analysis	First 12 PCs retained	78.9	77.8	78.4	0.81	64.5	85.2	74.1	0.79
	Yes	Pearson’s correlation coefficient	First 16 features: Cd Hg Zn Sb Pb Ag Mg Ba Fe S Ca Sr Cu As Mn V	77.5	69.8	73.9	0.79	71.0	81.5	75.9	0.75
	Yes	Correlation-based feature selection	11 features: Ag As Ba Ca Cu Hg Mg S Sb V Zn	73.2	69.8	71.6	0.77	64.5	81.5	72.4	0.77
	Yes	Information gain ratio	First 12 features: As Zn Cd Ag Hg Mg Pb Sb S Cu Ba Ca	76.1	65.1	70.9	0.77	67.7	81.5	74.1	0.77
	Yes	Wrapper–forward feature selection	4 features: S Sb Sr V	70.6	68.3	68.0	0.75	64.5	70.4	67.2	0.75
	Yes	Wrapper–backward feature elimination	16 features: Ag Al As Ba Ca Co Cr Cu Fe Hg Mg P Pb S Sr V	73.2	69.8	71.6	0.78	71.0	74.1	72.4	0.79

The original article has been corrected.