Skip to main content
. 2022 Feb 12;23(2):bbab518. doi: 10.1093/bib/bbab518

Figure 1.

Figure 1

For each of the six datasets, we optimized the hyperparameter space of 21 classifiers using 5-fold cross-validation on the training set. The predictions of the models with the best F1 score from these classifiers were provided as input to meta-classifiers. Given the F2 score on the validation set, the best meta-classifier was kept as the final predictor.