TabSeq Feature Ordering

This module extracts and packages the feature ordering algorithm used in TabSeq (ICPR 2024) as a standalone utility, enabling integration into any tabular deep learning pipeline.

Key Features

• Variance-based intra-cluster ordering
• KMeans clustering for feature grouping
• Weighted global ordering from local cluster orders
• Minimal dependencies, flexible integration

Installation

pip install tabseq-feature-ordering

Parameters

• X_train: Tabular training data as pd.DataFrame
• cluster_size: Number of clusters (e.g., 5)
• sort_order: Intra-cluster sorting order by variance ('ascending' or 'descending')

Output

• global_ordering: List of column names in reordered order
• X_train_reordered: DataFrame with reordered columns

Example

import pandas as pd
import numpy as np
from tabseq_feature_ordering import reorder_features

# Example input
X = pd.DataFrame(np.random.rand(40, 80), columns=[f"F{i}" for i in range(80)])

# Run feature ordering (descending = high variance first)
order, X_reordered = reorder_features(X, cluster_size=5, ascending=False)

# Print the full reordered feature list
print("Reordered Features:\n", order)

License

MIT License © 2024 Zadid Habib

Citation

If you use this module, please cite our paper:

Habib, Al Zadid Sultan Bin, Kesheng Wang, Mary-Anne Hartley, Gianfranco Doretto, and Donald A. Adjeroh. "TabSeq: A Framework for Deep Learning on Tabular Data via Sequential Ordering." In International Conference on Pattern Recognition, pp. 418-434. Cham: Springer Nature Switzerland, 2024.

Bibtex

@inproceedings{habib2024tabseq,
  title={TabSeq: A Framework for Deep Learning on Tabular Data via Sequential Ordering},
  author={Habib, Al Zadid Sultan Bin and Wang, Kesheng and Hartley, Mary-Anne and Doretto, Gianfranco and A. Adjeroh, Donald},
  booktitle={International Conference on Pattern Recognition},
  pages={418--434},
  year={2024},
  organization={Springer}
}