Skip to main content
. 2018 Apr 26;14(4):e1006106. doi: 10.1371/journal.pcbi.1006106

Table 2. Evaluation of RIDDLE and baseline classification methods.

All values are averaged over ten k-fold cross-validation experiments. In addition, the precision, recall and ROC scores are averaged across classes, weighted by the number of samples in each class. Support vector machines (SVMs) could not be evaluated on the full dataset as individual trials required more than 36 hours of computation. For runtime comparisons a standard computing configuration was used: 16 Intel Sandybridge cores at 2.6 GHz and 16GB RAM; graphics processing units were not utilized.

Method Accuracy Loss Precision Recall F1 Macro-average ROC Runtime (h)
RIDDLE 0.668 0.857 0.663 0.668 0.652 0.833 0.962
logistic regression 0.644 0.928 0.639 0.644 0.611 0.807 0.024
random forest 0.629 0.962 0.641 0.629 0.578 0.799 2.395
GBDT 0.634 0.948 0.635 0.634 0.592 0.793 0.265
SVM, linear kernel N/A N/A N/A N/A N/A N/A >36
SVM, polynomial kernel N/A N/A N/A N/A N/A N/A >36
SVM, RBF kernel N/A N/A N/A N/A N/A N/A >36