If you face imbalance data in your machine learning project, this package is here to pre-process your data. It is an efficient and ready-to-use implementation of MGS-GRF, an oversampling strategy designed to handle large-scale and mixed imbalanced data-set — with both continuous and categorical features.

The MGS-GRF package is supported by two peer-reviewed publications:

    A paper at AISTATS 2026 introducing the algorithm for continuous input: Do we need rebalancing strategies? A theoretical and empirical study around SMOTE and its variants 📄

    A paper at ECML-PKDD 2025 presenting the extension of MG to categorical features (MGS-GRF): Harnessing Mixed Features for Imbalance Data Oversampling: Application to Bank Customers Scoring 📄

🛠 Installation

From Pypi:

pip install mgs-grf

From source:

git clone git@github.com:artefactory/mgs-grf.git

And install the required packages into your environment (conda, mamba or pip):

pip install -r requirements.txt

💻 How to use the MGS-GRF Algorithm to learn on imbalanced data

Here is a short example on how to use MGS-GRF:

from mgs_grf import MGSGRFOverSampler

## Apply MGS-GRF procedure to oversample the data
mgs_grf = MGSGRFOverSampler(categorical_features=categorical_features, random_state=0)
X_train_balanced, y_train_balanced = mgs_grf.fit_resample(X_train_imbalanced, y_train_imbalanced)

## Encode the categorical variables
enc = OneHotEncoder(handle_unknown="ignore", sparse_output=False)
X_train_balanced_enc = np.hstack((X_train_balanced[:,numeric_features],
                                  enc.fit_transform(X_train_balanced[:,categorical_features])))
X_test_enc = np.hstack((X_test[:,numeric_features], enc.transform(X_test[:,categorical_features])))

# Fit the final classifier on the augmented data
clf = lgb.LGBMClassifier(n_estimators=100, verbosity=-1, random_state=0)
clf.fit(X_train_balanced_enc, y_train_balanced)

A more detailed notebook example is available in this notebook.

🙏 Acknowledgements

This work was done through a partenership between Artefact Research Center and the Laboratoire de Probabilités Statistiques et Modélisation (LPSM) of Sorbonne University.

📜 Citation

If you find the code useful, please consider citing us :

@inproceedings{sakho2024we,
  title={Do we need rebalancing strategies? A theoretical and empirical study around SMOTE and its variants},
  author={Sakho, Abdoulaye and Malherbe, Emmanuel and Scornet, Erwan},
  booktitle={The 29th International Conference on Artificial Intelligence and Statistics},
  year={2026},
  organization={PMLR}
}

@inproceedings{sakho2025harnessing,
  title={Harnessing Mixed Features for Imbalance Data Oversampling: Application to Bank Customers Scoring},
  author={Sakho, Abdoulaye and Malherbe, Emmanuel and Gauthier, Carl-Erik and Scornet, Erwan},
  booktitle={Joint European Conference on Machine Learning and Knowledge Discovery in Databases},
  pages={247--264},
  year={2025},
  organization={Springer}
}