The L-Curve Criterion as a Model Selection Tool in PLS Regression

Table 4

The result of the comparison between cross validation (CV), generalized cross validation (GCV), and L-curve criterion in PLS regression in terms of mean squared error of both the estimator and the predicted y.

Parameters	CV		GCV		L-curve
	MSE of coefficient	MSE of y	MSE of coefficient	MSE of y	MSE of coefficient	MSE of y
(m, 0.5, 1)	108	217780	13414	63990	6786	80820
(m, 0.5, 5)	0.0445	433.20	25.31	117.58	11.00	149.5405
(m, 0.5, 7)	0.0414	112.4997	6.43	29.0744	2.90	38.8957
(m, 0.8, 1)	97	559670	8679.3	151260	3596	199000
(m, 0.8, 5)	0.435	1000.9	135.98	269.7	64.43	343
(m, 0.8, 7)	0.0599	380.83	38.46	148.54	19.94	169.22