SMOTE (Synthetic Minority Over-sampling)

What is SMOTE (Synthetic Minority Over-sampling)?

A technique to handle imbalanced datasets (e.g., fraud detection). Instead of just duplicating minority samples, it synthesizes new, plausible examples by interpolating between existing ones.

Where did the term "SMOTE (Synthetic Minority Over-sampling)" come from?

Chawla et al. (2002).

How is "SMOTE (Synthetic Minority Over-sampling)" used today?

Go-to fix for 'rare event' prediction problems.

Related Terms

• imbalanced-data
• data-augmentation