Fig. 10

Depicts 5XCV outcomes on three learning machines used in the optimal crowd for Breast Cancer Data. 200 case and 200 control samples were used for the 5XCV. The percent errors of each fold were averaged for a final result. The error bars represent one standard error. The label “SYN = True” means that synthetic features were used in the optimal crowd, and the label “SYN = False” refers to the lack of synthetic features. Panel a depicts machine performances for classification based on probability, and Panel b depicts machine performances for pure classification