Skip to main content
Fig. 6 | BMC Bioinformatics

Fig. 6

From: Cost-Constrained feature selection in binary classification: adaptations for greedy forward selection and genetic algorithms

Fig. 6

Performance results for the plasmode simulation Setting R and the real world data Setting S. Boxplots for every feature selection method illustrate the distribution of the AUC values obtained for the 100 training-test splits (transparent dots). The black diamonds depict the mean AUC values. A green bar in the top plot highlights the area between the 0.05 and 0.95 quantile of AUC values when always selecting the optimal subset of relevant features that fit in the budget. For Setting S, in the bottom plot, the left elements show the results with cmax=1.5 and the right elements show the results with cmax=3

Back to article page