A collection of various oversampling techniques developed from SMOTE is provided. SMOTE is a oversampling technique which synthesizes a new minority instance between a pair of one minority instance and one of its K nearest neighbor. (see <https://www.jair.org/media/953/live-953-2037-jair.pdf> for more information) Other techniques adopt this concept with other criteria in order to generate balanced dataset for class imbalance problem.
copied from cf-staging / r-smotefamily