Fig. (7).

Performance of ELM and RF holdouts at the training and testing stages. It is possible to observe the remarkable difference in the performance of both classifiers, where the inherent robustness of RF significantly outperforms ELM. The testing median accuracy of ELM is around 0.6, while for RF, it is greater than 0.9. Therefore, the holdout RF algorithm clearly outperforms the ELM holdouts.