Skip to main content

> ML_LIBRARY // IMBALANCED-LEARN_v1.0

imbalanced-learn

scikit-learn-contrib — Toolkit for classification with severely imbalanced datasets.

classical-mlv0.12.4MITqualified
Official DocumentationSource Repository

Model Training

Supported
Accelerators:
CPU
Distributed Training:No

Model Inference

Supported
Inference Accelerators:
CPU
Deployment Targets:server, edge

What It Does

  • +Over-sampling using synthetic minority instances (SMOTE, ADASYN)
  • +Under-sampling of majority classes (RandomUnderSampler, TomekLinks, NearMiss)
  • +Balanced ensemble classifiers compatible with scikit-learn

What It Does Not Do

  • -Natively accelerate on GPUs
  • -Directly handle streaming time series drift
  • -Train deep generative networks

>Suitable Work Types

  • Financial fraud detection with <0.1% positive class prevalence
  • Rare medical condition diagnosis
  • Manufacturing defect detection

>Unsuitable Work Types

  • Balanced multiclass problems
  • Computer vision generative augmentation
Data Residency Implications

In-process memory only.

Security Considerations

Safe library; standard scikit-learn pipeline security applies.

Operational Profile & Known Limitations

Maturity:mature
Learning Curve:low
Ops Complexity:low
Cost Tier:free-oss
> Known Limitations:
  • Synthetic sample generation can cause synthetic data leakage if applied before cross-validation split.
  • High computational complexity on large datasets with many dimensions.

Associated Incident Patterns (Incidentpedia)

Enforce safeguards and monitoring to guard against these documented real-world failure modes:

training-serving-skew

> Primary Evidence & Benchmark Citations

imbalanced-learn Documentationofficial-docs • >=0.10.0, <=0.12.x
2026-09-25HIGH