bias-detection-engine 0.1.0

A tool for detecting biases in machine learning datasets

Bias Detection Engine for ML Datasets

A comprehensive tool for detecting various types of biases in machine learning datasets before model training begins.

Features

• Predictive Imbalance Detection: Identifies class imbalance and feature distribution skews
• Label Leakage Analysis: Detects potential data leakage and feature dependencies
• Distributional Bias: Analyzes demographic and categorical feature distributions
• Information Theoretic Metrics: Measures feature importance and mutual information
• Autoencoder-based Anomaly Detection: Identifies unusual patterns in the data
• Contrastive Learning Analysis: Evaluates feature representations and relationships

Installation

pip install -r requirements.txt

Project Structure

bias_detection_engine/
├── core/
│   ├── __init__.py
│   ├── imbalance_detector.py
│   ├── leakage_detector.py
│   ├── distribution_analyzer.py
│   └── feature_analyzer.py
├── utils/
│   ├── __init__.py
│   ├── preprocessing.py
│   └── visualization.py
├── models/
│   ├── __init__.py
│   ├── autoencoder.py
│   └── contrastive_learner.py
├── notebooks/
│   └── examples/
└── tests/
    └── __init__.py

Usage

from bias_detection_engine.core import ImbalanceDetector, LeakageDetector, DistributionAnalyzer
from bias_detection_engine.utils import preprocess_data

# Load and preprocess your dataset
data = preprocess_data(your_dataset)

# Detect imbalances
imbalance_detector = ImbalanceDetector()
imbalance_report = imbalance_detector.analyze(data)

# Check for label leakage
leakage_detector = LeakageDetector()
leakage_report = leakage_detector.analyze(data)

# Analyze distributions
distribution_analyzer = DistributionAnalyzer()
distribution_report = distribution_analyzer.analyze(data)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.