FairSample

A Python toolkit for handling imbalanced datasets with 14+ resampling techniques and 40+ complexity measures.

Why This Package?

Most imbalanced learning packages only provide resampling techniques. This toolkit adds complexity measures to help you:

• Understand why your dataset is difficult
• Identify class overlap and boundary issues
• Choose the best technique for your data
• Measure improvement after resampling

Key Features

• 14+ Resampling Techniques - Overlap-based, hybrid, and clustering methods
• 40+ Complexity Measures - Feature, instance, structural, and multiresolution metrics
• Scikit-learn Compatible - Standard fit_resample() API
• Pandas Support - Works seamlessly with DataFrames
• No Forced Workflow - You control training and evaluation

Quick Example

from fairsample import RFCL
from fairsample.complexity import ComplexityMeasures

# Check complexity
cm = ComplexityMeasures(X, y)
print(f"Overlap: {cm.analyze_overlap()['N3']:.4f}")

# Apply resampling
sampler = RFCL(random_state=42)
X_resampled, y_resampled = sampler.fit_resample(X, y)

What's Included

Resampling Techniques

• Overlap-Based: RFCL, NUS, URNS, DeviOCSVM, FCMBoostOBU
• Hybrid: SVDDWSMOTE, ODBOT, EHSO
• Clustering: NBUS (4 variants), KMeansUndersampling (4 variants)
• Comprehensive: OSM
• Baselines: RandomOverSampler, RandomUnderSampler

Complexity Measures

• Feature Overlap (6): F1, F1v, F2, F3, F4, Input Noise
• Instance Overlap (9): N3, N4, kDN, CM, R-value, D3, SI, Borderline, Degree of Overlap
• Structural (9): N1, N2, T1, DBC, LSC, Clust, NSG, ICSV, ONB
• Multiresolution (5): Purity, Neighbourhood Separability, MRCA, C1, C2

Get Started

Installation Quick Start Examples

License

MIT License - Free for commercial and personal use.