Skip to main content

Table 3 Five-fold cross-validations of meta classifiers

From: A meta-learning approach for B-cell conformational epitope prediction

Classifier TPR FPR Precision Accuracy F-score MCC AUC
2-level (ANN)a 0.514 0.019 0.705 0.944 0.594 0.573 0.748
2-level (C4.5)a 0.511 0.023 0.663 0.941 0.577 0.551 0.744
2-level (k-NN)a 0.496 0.012 0.783 0.949 0.607 0.599 0.742
2-level (SVM)a 0.593 0.009 0.848 0.959 0.697 0.689d 0.920
3-level Stackingb 0.579 0.009 0.850 0.958 0.689 0.682d 0.925
Cascadec 0.588 0.010 0.843 0.959 0.693 0.684d 0.925
  1. aTwo-level stacking meta classifiers with ANN, C4.5, k-NN, or SVM as the top-level meta learner.
  2. bThree-level stacking meta classifier (Figure 3).
  3. cCascade meta classifier (Figure 4).
  4. dPaired t test showed no significant difference.