Comparison of predictive quality on the BCPREDFrag dataset calculated using 10-fold cross validation. The methods are sorted by their AUC values in the ascending order.

aresults from Table 1 in [21].bresults from Table II in [22].cresults for the SVM model (C = 8.0 and gamma = 0.000977) that uses all 198 features.dresults for the SVM model (C = 1.0 and gamma = 0.001953) that uses the selected 84 features.