Figure 2.

Comparison of accuracy and F-measure in datasets with different positive-to-negative ratios. The Random predictor predicts any query protein pair as positive with a probability of 0.5, and as negative with a probability of 0.5, too. The Opportunistic predictor predicts any query protein pair as negative for accuracy and it predicts any query protein pair as positive for F-measure. Shen et al. use triad frequency as the feature set. Guo et al. use auto cross covariance as the feature set. This work uses triad significance as the feature set.