Leakage: SMOTE before the split

Show in practice that applying SMOTE before splitting inflates results.

AdvancedData Balance Lab
العربية

The idea

The lab compares the correct pipeline (split, then SMOTE on training) with the wrong one.

Data description

Synthetic data generated in your browser from stated distributions; no real personal data.

Suggested next experiment

Concepts in this experiment

Data leakage

Using test information during training or preprocessing, making results look better than they are, e.g. SMOTE, scaling or feature selection before the split.

Scientific sources for this experiment

  1. SMOTE: Synthetic Minority Over-sampling Technique (arXiv:1106.1813)N. V. Chawla, K. W. Bowyer, L. O. Hall, W. P. Kegelmeyer (2002) — Paper
  2. imbalanced-learn — SMOTEimbalanced-learn developers (2025) — Tool

More in this module