Fig. 2From: Joint use of over- and under-sampling techniques and cross-validation for the development and assessment of prediction modelsProbability that at least one of the replicas of a sample included in the test fold is included also in the training fold, as a function of the proportion of minority class samples (p min ). The figure shows how the probability that a test sample has a replica in the learning fold depends on the level of class-imbalance (p m i n ) in a dataset with n=100 samples when using 2 fold CVBack to article page