Table 2.

Area under curve (AUC) and classification accuracy (CA) for machine learning models at yield cutoff of 50 Mg ha⁻¹ using raw concentrations or clr values as features.

Learner	Nutrient concentration values		Centered log-ratio-transformed nutrient concentration values
	Area under curve	Accuracy	Area under curve	Accuracy
Random forest	0.994	0.940	0.996	0.959
Extreme gradient boosting	0.993	0.928	0.996	0.945

N, P, K, Ca, Mg, Fe, Mn, Zn, Cu, B, S.