Quantification of the Spectral Variability of Ore-Bearing Granodiorite under Supervised and Semisupervised Conditions: An Upscaling Approach

Table 7

The prediction results using the kNN method for each level.

		Observed	Predicted			Correct (%)
			M1	M2	M3
L0 (N = 410)	Training	M1	50	1	6	87.7
		M2	1	71	15	81.6
		M3	5	34	89	69.5
		Overall (%)	20.6	39.0	40.4	77.2
	Test	M1	32	0	1	97.0
		M2	0	21	12	63.6
		M3	3	28	41	56.9
		Missing	0	0	0	N.D.^§
		Overall (%)	25.4	35.5	39.1	68.1
L1 (N = 41)	Training	M1	2	0	4	33.3
		M2	0	5	5	50.0
		M3	1	2	10	76.9
		Overall (%)	10.3	24.1	65.5	58.6
	Test	M1	3	0	0	100
		M2	0	0	2	0
		M3	0	2	5	71.4
		Missing	0	0	0	N.D.^§
		Overall (%)	25.0	16.7	58.3	66.7
L2 (N = 12)	Training	M1	0	1	1	0
		M2	1	1	1	33.3
		M3	1	1	0	0
		Overall (%)	28.6	42.9	28.6	14.3
	Test	M1	0	0	1	0
		M2	0	0	1	0
		M3	0	3	0	0
		Missing	0	0	0	N.D.^§
		Overall (%)	0.0	60.0	40.0	0
L0x (N = 90)	Training	M1	20	0	0	100
		M2	0	21	1	95.5
		M3	0	0	19	100
		Overall (%)	32.8	34.4	32.8	98.4
	Test	M1	10	0	0	100
		M2	0	8	0	100
		M3	0	0	11	100
		Missing	0	0	0	N.D.^§
		Overall (%)	34.5	27.6	37.9	100.0

§No data.