Table 3.

Confusion matrix of SVC models using extracted features

Cultivar	Parametera		Training set					Validation set					Testing set
			CK	Q	S	Accuracy	Kappa	CK	Q	S	Total	Kappa	CK	Q	S	Accuracy	Kappa
XS 134	(105, 10–1)	CK	25	0	0			6	0	0			5	0	1
		Q	0	21	5			0	6	0			0	6	1
		S	0	2	26			0	2	5			1	0	6
						91.14%	86.68%				89.47%	80.65%				85%	76.15%
ZJ 88	(106, 10–1)	CK	28	0	0			7	0	0			6	0	0
		Q	0	28	0			0	6	1			0	6	0
		S	0	0	26			0	0	6			0	2	5
						100%	100%				95%	92.5%				89.47%	80.5%

^aParameter means the model parameter of SVC, which is the combination of penalty coefficient C and RBF kernel parameter g, i.e. (C, g)

^bTotal means the total classification accuracy

^cKappa is used to evaluate the inter-rater reliability of the classification results

CK: treated with nutrient solution, Q: treated with 0.25 g/L quinclorac, S: pre-treated with 10 mg/L SA followed by 0.25 g/L quinclorac