Table 2.

The results of the models built by three algorithms (PLS and EN-PLS).

Model	# of variables	Optimum # of PLS components	Training set			Test set
			R t 2	RMSET	RMSECV	R p 2	RMSEP
PLS	10493	8	0.87	5.95	16.63	0.34	12.70
EN-PLS	309	7	0.93	4.34	6.93	0.55	11.66

R ² is correlation coefficient of regression between the predicted and experimental activities of the extracts (t refers to training set and p refers to the test set); RMSET is the fitting error of the model in the training; RMSECV is the Root Mean Squared Errors of Cross-Validation; RMSEP is Root Mean Squared Errors of Prediction of the test set; q² is the cross-validated R² which is calculated by the equation: q² = 1 − ∑(Y_pred − Y_act)²/∑(Y_act − Y_mean)².