Average performance metrics, per level of class imbalance, across training, validation, and held-out test sets for all class-balancing ensembles and all routes/modes.

The average performance is calculated as the mean over 50 iterations for the training, validation, and test sets, and over 10 iterations for the top-10 ensembles. Except for ROC-AUC and PR-AUC, all other metrics were computed at >0.5 probability threshold. Brier scores range from 0 (best performance) to 1 (worst performance), while MCC values range from +1 (best performance) to -1 (worst performance). ± values indicate standard deviation from the mean. Values in square brackets indicate the worst and best performing ensembles, respectively. Fig AA in S1 Text visualises difference in performance between held-out test sets and validation sets, per level of class imbalance, for all class-balancing ensembles and their constituent models. Fig AB in S1 Text illustrates variance in performance on the test-sets across the 50 iterations, per level of class imbalance, for all class-balancing ensembles and their constituent models.