A Progressive Combined Variable Selection Method for Near-Infrared Spectral Analysis Based on Three-Step Hybrid Strategy

Table 1

Statistical results of oil and starch content in corn samples and organic matter content in soil samples.

Substance	Sample set	Number of samples	Minimum	Maximum	Mean	Standard deviation
Oil in corn	Training set	60	3.088	3.832	3.482	0.187
	Test set	20	3.264	3.766	3.547	0.136
	Total set	80	3.088	3.832	3.498	0.177
Starch in corn	Training set	60	62.826	66.472	64.737	0.812
	Test set	20	63.021	65.808	64.571	0.856
	Total set	80	62.826	66.472	64.696	0.821
Organic matter in soil	Training set	81	42.910	95.850	84.768	11.853
	Test set	27	64.610	93.010	87.408	6.656
	Total set	108	42.910	95.850	85.428	10.822

The training set and the test set were divided as 3:1 ratio by Kennard–Stone algorithm.