Fig. 7From: A novel generative adversarial networks modelling for the class imbalance problem in high dimensional omics dataFeatures Selected Microarray Data Validation Performance. Validation scores are summarised across experiments using the microarray (LASSO and EN feature selection) data for GAN, ExpandGAN, SMOTE and RO methods. Boxes represent 1 standard deviation from the mean, and the horizontal line, the mean. The ‘AUC Diff’ defines the score difference between the 0.5 baseline (red, separated lines) and the validation score. Adjacent, same-coloured bars define results using different alpha hyperparameters, in order of 0/0, 1/0, 1/1 for ExpandGAN and 0 and 1 for GAN. Where/if SMOTE varied across experiments with changed alpha, the results are shown in multiple adjacent bars Area under the receiver operating characteristic curve (AUC).Back to article page