ML internals: Synthetic Minority Oversampling (SMOTE) Technique
Domino Data Lab
MAY 20, 2021
Other techniques include simple re-sampling, where the minority class is continuously re-sampled until the number of obtained observations matches the size of the majority class, and focused under-sampling, where the discarded observations from the majority class are carefully selected to be away from the decision boundary (Japkowicz, 2000).
Let's personalize your content