Figure 2

(a) Performance of prediction models with no DR method applied to the spectral data. (b) Performance of prediction models with DR methods applied to the spectral data. Overview of the prediction performance of ML models for PL-based sample identification. The accuracy and the \(f_1\) score are presented as box plots and are calculated during the validation stage. The colored boxes show the quartiles of the achieved scores, while the whiskers extend to show the rest of the distribution. Each plot presents the prediction performance with a different classifier. All plots in (a) show the performance when the spectral data has not been processed with a DR method. All models in the plots in (b) are built around spectral data that have been processed with either PCA or SDCM. The prediction performance at all training stages are summarized in Fig. S1.