Fig. 5

Classification assessment of seventeen genes associated with triple-negative breast cancer recurrence in chemoresistant tumors. Six classifiers were evaluated with the SuperLearner package employing seventeen genes and three downs-selection methods. The mean cross-validation AUC is plotted as a point for each classification method, and the four feature selection methods used are shown. We observed that three models, kernel support vector machine, kNN, and random forest models, achieved the best overall median classification AUC