Table 4.

Best fit models for estimating the percentage of cones damaged by Megastigmus albifrons, based on the sample of 192 Pinus strobiformis trees.

Method of variable selection	Machine learning algorithm	Independent variables	RMSE	MAE	R2
PLS	brnn	%Organic matter, Znppm, KCEC, MTWM, Reg P. strobiformis, Cuppm, Mnppm, HC	0.0085	0.0061	0.983
PLS	rf	%Organic matter, Znppm, KCEC, MTWM, Reg P. strobiformis, Cuppm, Mnppm, HC	0.0100	0.0061	0.993
PLS	avNNet	%Organic matter, Znppm, KCEC, MTWM, Reg P. strobiformis, Cuppm, Mnppm, HC	0.0800	0.0691	0.454
PLS	nnet	%Organic matter, Znppm, KCEC, MTWM, Reg P. strobiformis, Cuppm, Mnppm, HC	0.0805	0.0693	0.462
PLS	mlpWeightDecay	%Organic matter, Znppm, KCEC, MTWM, Reg P. strobiformis, Cuppm, Mnppm, HC	0.0833	0.059	0.439
ROC	lm	CEC, MAP, Reg P. strobiformis, Mnppm, MMAX, %Organic matter, SMRPB, MgCEC	0.0834	0.0689	0.396

PLS, Partial Least Squares; ROC, Receiver Operating Characteristic; brnn, Bayesian Regularized Neural Networks; rf, Random Forest; avNNet, Model Averaged Neural Network; nnet, Neural Network; mlpWeightDecay, Multi-Layer Perceptron; lm, linear regression; RMSE, Root-mean-square error; MAE, Mean Absolute Error; R², Determination coefficient; %Organic matter, Relative proportion of organic matter in CEC (%) in the soil; Zn_ppm, zinc concentration in the soil (ppm); K_CEC, proportion of potassium in CEC (cation exchange capacity); MTWM, Mean temperature in the warmest month (degrees C); Reg P. strobiformis, Pinus strobiformis regeneration; Cu_ppm, copper concentration in the soil (ppm); Mn_ppm, manganese concentration in the soil (ppm); HC, Hydraulic conductivity (cm/h); CEC, Cation exchange capacity; MAP, Mean annual precipitation (mm); MMAX, Mean maximum temperature in the warmest month (degrees C); SMRPB, Summer precipitation balance: (jul+aug+sep)/(apr+may+jun); Mg_CEC, proportion of magnesium in CEC (cation exchange capacity).