Chain based sampling for monotonic imbalanced classification.

•We designed and implemented a novel sampling scheme based on monotonic chains in order to preverse monotonicity while obtaining a balanced set.•This scheme has been included in five famous under- and over-sampling approaches: Random Under-Sampling, Random Over-Sampling, SMOTE, ADASYN and MWMOTE.•Empirically, we show the deterioration of the monotonicity degree in 8 data-sets caused by standard and ordinal sampling approaches, invalidating their use for monotonic classification.•We compare our new monotonic sampling techniques versus their standard versions, achieving a better monotonicity preservation than the standard ones with the same improvement of predictability.