SMOTE and data augmentation techniques

[briannaburton]

http://contrib.scikit-learn.org/imbalanced-learn/stable/

[briannaburton]

Random over sampling to generate new samples in the under-represented class - 3 different types = 3 different decision functions

• RandomOverSampler: over-sampling by duplicating some of the original samples of the minority class
• Synthetic Minority Oversampling Technique (SMOTE): generate new samples in by interpolation and will not make any distinction between easy and hard samples to be classified using the nearest neighbors rule
• Kind parameter affects how border of the decision function handled
• Adaptive Synthetic (ADASYN) sampling method: generating samples next to the original samples which are wrongly classified using a k-Nearest Neighbors classifier

[briannaburton]

SMOTEEN and SMOTETomek combination of over and under sampling

• Smoteen over-sampling using SMOTE and cleaning using ENN.
• Smotetomek over-sampling using SMOTE and cleaning using Tomek links