Skip to main content

Table 2 Comparison of solutions to the data imbalance issue on the edge detection dataset.

From: Self-training in significance space of support vectors for imbalanced biomedical event data

Class label Precision (%) Recall (%) F-measure (%)
  Under-sampling Class weighting STSS Under-sampling Class weighting STSS Under-sampling Class weighting STSS
AtLoc 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Cause 100.0 42.24 47.36 4.48 25.95 43.67 8.51 32.14 45.44
Cause-Theme 100.0 63.07 76.4 40.47 63.32 50.21 57.62 63.19 60.59
Site 0.0 20.78 29.3 0.0 12.37 33.21 0.0 15.5 31.13
Theme 53.87 60.91 57.16 42.21 62.37 90.83 47.33 61.63 70.16
ToLoc 26.91 57.39 48.2 99.74 54.33 96.11 42.38 55.81 64.2
Negative 89.66 92.35 100.0 100.0 100.0 100.0 94.54 96.02 100.0
\