SheShe

📚 Full documentation: https://jcval94.github.io/SheShe/

Smart High-dimensional Edge Segmentation & Hyperboundary Explorer

SheShe turns probabilistic models into guided explorers of their decision surfaces, revealing human‑readable regions by following local maxima of class probability or predicted value.

Features

• Supervised clustering for classification and regression
• Rule extraction and subspace exploration
• 2D/3D plotting utilities

Mathematical Overview

SheShe approximates the optimisation problem max_x f(x) by climbing gradient-ascent paths toward local maxima and delineating neighbourhoods around them. Detailed derivations for each module are provided in the documentation.

Installation

Requires Python ≥3.9.

Dependencies

Main

• NumPy for numerical operations
• Pandas for tabular data
• scikit-learn for model training
• Matplotlib for visualisation
• hnswlib for fast nearest-neighbour search

Optional

• LightGBM for gradient boosting models
• SHAP and Interpret for model explanations

From PyPI

Install from PyPI:

pip install sheshe

From source

git clone https://github.com/jcval94/SheShe.git
cd SheShe
pip install -e .

Common issues

• Windows: Compiling packages such as hnswlib or lightgbm may require the Build Tools for Visual Studio. Alternatively, use a conda environment or install via WSL.
• macOS: Ensure Xcode command-line tools are installed (xcode-select --install). Some dependencies (e.g. lightgbm) also need OpenMP support: brew install libomp. If wheels are unavailable, build from source using Homebrew-provided compilers.

Documentation

See the documentation for installation, API reference and guides.

Contributing

Set up a virtual environment and install the development dependencies:

pip install -r requirements-dev.txt
pip install -e .

Run the tests to ensure everything works:

pytest

No linter is currently configured; feel free to run black . locally before submitting changes.

Author

SheShe is authored by José Carlos Del Valle – LinkedIn | Portfolio

License

MIT