Table 1.
Resampling methods.
Fig 1.
Visual representation of SMOTE.
xi: Randomly selected minority class sample; xzi: Instance close to xi; xnew: New artificial example generated by interpolation between two instances.
Fig 2.
Illustration of TL.
Fig 3.
Backbone model under low complexity (c = 1).
MIN: Minority class; MAJ: Majority class.
Fig 4.
Backbone model under medium complexity (c = 2).
Fig 5.
Backbone model under extreme complexity (c =2, but classes are spaced apart).
Table 2.
Simulation settings.
Fig 6.
Differences between ROC and PR curves.
Fig 7.
Differences in AUPRC ranks under low complexity.
The star markers at the top of bars indicate significant performance gains.
Fig 8.
Differences in AUPRC ranks under medium complexity.
The star markers at the top of bars indicate significant performance gains.
Fig 9.
Difference in AUPRC ranks under extreme complexity.
The star markers at the top of bars indicate significant performance gains.
Table 3.
Summary of complexity measures.
Fig 10.
Mean differences in AUPRC values using F3.
The more complex the dataset, the darker the color of the bar.
Fig 11.
Mean differences in AUPRC values using N2.
The more complex the dataset, the darker the color of the bar.
Fig 12.
Mean differences in AUPRC values using C2.
The more complex the dataset, the darker the color of the bar.
Fig 13.
Complex and noncomplex areas in real datasets.
Table 4.
Top 10 combinations for complex and noncomplex datasets obtained using F3.
Table 5.
Top 10 combinations for complex and noncomplex datasets obtained using N2.