Quantitative Detection of Quartz Sandstone SiO₂ Grade Using Polarized Infrared Absorption Spectroscopy with Convolutional Neural Network Model

Table 5

Assessment of the GRNN prediction results.

Input variable	R^2	RMSE	RPIQ
Original data	0.3932	0.061	1.93
First-order derivative	0.5983	0.045	2.61
Reciprocal logarithm	0.5388	0.051	2.30
MSC	0.8722	0.026	4.52
Original data-PCA	0.8844	0.025	4.70
First-order derivative-PCA	0.6741	0.04	2.94
Reciprocal logarithm-PCA	0.7917	0.034	3.46
MSC-PCA	0.8284	0.029	4.05
Original data-SPA	0.4422	0.057	2.06
First-order derivative-SPA	0.6592	0.041	2.87
Reciprocal logarithm-SPA	0.5222	0.052	2.26
MSC-SPA	0.8646	0.026	4.52