Figure 3. Dataset effect on model performance.

Accuracy distribution at 100 fold MCCV for the eight positive datasets employed in this study (see Fig. 4 for negative datasets). All other variables were summarized (e.g., six learners).