2.4.1. Wavelength Extraction Method

The SPA [35,36] eliminates the influence of collinearity among various variables, reduces the overlap of effective information, and accelerates the modeling speed.

The IRIV method [37] utilizes random combinations of variables and interactions between variables to select variables based on binary matrix rearrangement filters. Based on the model cluster analysis method, the difference of mean values (DMEAN) and the P value were calculated to determine the class of each variable. The classification rules of variables are shown in Table 2.

**Table 2.** Variable classification rules of iteratively retains informative variables (IRIV) [37].

