A Variable Selection Method Based on Fast Nondominated Sorting Genetic Algorithm for Qualitative Discrimination of Near Infrared Spectroscopy

Table 2

Classification results of PLS-DA model based on different variables selection methods.

Model	Number of variables	Class	Calibration set		Prediction set
			Sen	Spe	CDR (%)	RDR (%)	Sen	Spe	CDR (%)	RDR (%)
PLS-DA	1036	B	0.879	0.917	79.57	96.77	0.800	0.905	77.42	96.77
		C	0.667	0.905			0.583	0.947
		X	0.833	0.873			1.000	0.818
NSGA-II-PLS-DA	160	B	0.970	0.933	87.10	100	0.900	0.857	80.65	100
		C	0.633	0.984			0.583	0.947
		X	1.000	0.889			1.000	0.909
CARS-PLS-DA	91	B	0.970	0.917	84.95	98.92	1.000	0.857	80.65	100
		C	0.633	0.968			0.583	0.947
		X	0.933	0.889			0.889	0.909

Note “Sen,” “Spe,” “CDR,” and “RDR” denote sensitivity, specificity, correct discriminant rate, and reasonable discriminant rate, respectively.