2.4. Spectra data corrections

To achieve robust and accurate prediction results, spectra data need to be corrected and pre-processed prior to prediction model development. To compensate some noises on spectra data, smoothing with Savitsky-Golay algorithm can be applied to eliminate noises. Then, mean normalization (MN) can be also a choice as spectra correction method to remove background irrelevant information and normalize spectral data. Moreover, standard normal variate (SNV) is a good choice to pre-process and enhance spectra data as this method is widely employed beside multiplicative scatter correction (MSC) either as full MSC or as extended (EMSC). Both these methods proven to be effective and improve prediction performance.