shapash

🔅 Shapash: User-friendly Explainability and Interpretability to Develop Reliable and Transparent Machine Learning Models

What is shapash?

🔅 Shapash: User-friendly Explainability and Interpretability to Develop Reliable and Transparent Machine Learning Models

Overview of shapash

What it is

Shapash is a Python library for making machine learning models interpretable and comprehensible. It provides visualizations with explicit labels and can generate a webapp to navigate between global and local explainability. The library is compatible with models such as Catboost, Xgboost, LightGBM, Sklearn Ensemble, linear models, and SVM. It supports regression, binary classification, and multiclass problems. Shapash can use Shap or Lime backends to compute feature contributions and works with category-encoders and sklearn ColumnTransformer.

Key features

  • Displays clear results using explicit labels for features and their values.
  • Generates a webapp for navigating between global and local explainability.
  • Summarizes and exports local explanations.
  • Evaluates explainability quality with metrics: Stability, Consistency, and Compacity.
  • Allows filtering of subsets based on explanatory features, additional features, and prediction correctness.
  • Supports deployment of interpretability in API or batch mode.
  • Generates a standalone HTML report for model auditability.

Use cases

Shapash is used by data scientists to understand model behavior and share results with non-technical stakeholders. It aids in model auditing by producing a report that can be shared with internal audit, DPO, risk, and compliance teams. It can also be used to select and analyze specific subsets of data for deeper explainability.

Getting started

Installation can be done with pip:bashpip install shapash

License and pricing

Shapash is open source under the Apache-2.0 license.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/maifshapash.svg)](https://yunodb.com/tools/maifshapash?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/maifshapash?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/maifshapash.svg" alt="Featured on YunoDB" height="20"></a>