Table 2 Prediction precision and accuracy of chemical composition parameters (crude protein and neutral detergent fiber content) of Marandu palisadegrass pastures using random forest and support vector regression models.

Variables	Input features^#	Linear regression				CCC
	Input features^#	R²		RMSPE		CCC
	Model	RF	SVR	RF	SVR	RF	SVR
CP (g/g DM)	Bd	0.50 (0.07)	0.64 (0.06)	0.03 (0.002)	0.03 (0.002)	0.64 (0.05)	0.79 (0.04)
	Bd + Mt	0.58 (0.09)	0.66 (0.11)	0.03 (0.002)	0.03 (0.004)	0.70 (0.06)	0.80 (0.07)
	VI	0.48 (0.04)	0.61 (0.10)	0.03 (0.001)	0.03 (0.003)	0.63 (0.03)	0.77 (0.07)
	VI + Mt	0.51 (0.12)	0.65 (0.09)	0.03 (0.003)	0.03 (0.003)	0.66 (0.07)	0.80 (0.05)
	Bd + VI	0.55 (0.05)	0.58 (0.08)	0.03 (0.001)	0.03 (0.002)	0.67 (0.03)	0.75 (0.05)
	Bd + VI + Mt	0.57 (0.10)	0.64 (0.07)	0.03 (0.002)	0.03 (0.002)	0.70 (0.06)	0.80 (0.04)
NDF (g/g DM)	Bd	0.29 (0.07)	0.36 (0.08)	0.05 (0.004)	0.05 (0.003)	0.47 (0.07)	0.54 (0.06)
	Bd + Mt	0.49 (0.07)	0.57 (0.08)	0.04 (0.004)	0.04 (0.003)	0.65 (0.06)	0.73 (0.06)
	VI	0.27 (0.07)	0.39 (0.07)	0.05 (0.003)	0.05 (0.003)	0.45 (0.07)	0.58 (0.05)
	VI + Mt	0.47 (0.09)	0.56 (0.09)	0.04 (0.004)	0.04 (0.005)	0.64 (0.07)	0.73 (0.07)
	Bd + VI	0.33 (0.04)	0.40 (0.08)	0.05 (0.002)	0.05 (0.003)	0.48 (0.04)	0.59 (0.06)
	Bd + VI + Mt	0.49 (0.07)	0.57 (0.10)	0.04 (0.003)	0.04 (0.005)	0.65 (0.05)	0.73 (0.07)

The number in parenthesis represents a standard error among fivefold cross-validation. Bold represents the best models.
RMSPE root mean square prediction error, CCC concordance correlation coefficient, RF random forest, SVR support vector regression, CP crude protein, NDF neutral detergent fiber.
^#Bd: only data from spectral bands; see Table 3 for more information; Mt: meteorological data; maximum temperature (T_max), minimum temperature (T_min), average temperature (T_avg), relative humidity (RH), number of rainy days within a month (ND), rainfall; VI: only data from vegetation indices; see Table 4 for more information.

Search