Confusion matrix difference between the random forest classifiers trained on RAINBOW and MONOCHROMATIC features (normalized on precision). The dataset is composed of 300 light curves of each class (SNIa, SNII, SNIbc, SLSN, and TDE). The numbers represent the difference (RAINBOW – MONOCHROMATIC) in the median score of 100 iterations of bootstrapping. Individual confusion matrices for each method are given in Appendix B (Figs. B.1 and B.2).