Skip to main content
. 2022 Sep 19;17(9):e0274691. doi: 10.1371/journal.pone.0274691

Fig 2.

Fig 2

Consensus maps among each pair of the four classifiers, Random Forest (RF), Support Vector Machine (SVM), XGBoost (XGB) and Naïve Bayes (NB), (a) before and (b) after the cleaning up procedure applied for the 10-year IDE prediction. Consensus is measured by computing the Cohen’s kappa (κ) coefficient between each pair of classifiers on training sets and then averaged over 20 5-fold cross validation rounds.