MAIF/shapash 2.8.0 on GitHub

Major updates in this release:

WebApp Explainability Expansion: Introduction of several new visual explainability tools in the Shapash WebApp, including Global–Local Feature Importance, Clusters by Explainability, significantly enhancing model interpretation capabilities.
Modernized Packaging: Full support for numpy>=2.0.0 and dependency modernization, ensuring forward compatibility with the latest scientific Python ecosystem.

Add Global–Local Feature Importance Plot to WebApp
by @guillaume-vignal in #656
- Introduces a new visualization combining global feature importance with local (instance-level) contributions, bridging the gap between model-level and individual explanations directly in the WebApp.
Add Clusters by Explainability Plot Function and also to WebApp
by @guillaume-vignal, @Yh-Cherif in #658, #671, #632
- Provides a novel map-based visualization to explore and compare individuals based on their explanatory profiles.
- Enables projection of Shapley contributions into a reduced space, facilitating the identification of explanatory patterns and clusters across observations.
Add Column Ordering Support for Additional Data in WebApp
by @guillaume-vignal in #643
- Allows explicit control over the display order of additional data columns, improving readability and consistency in the WebApp.
Add _error_ Column Support for Classification
by @guillaume-vignal in [#663
- Adds explicit support for classification error tracking in datasets and visualizations.

Add cat_num_threshold Parameter to distribution_plot
by @guillaume-vignal in #646
- Improves automatic handling of categorical vs numerical features in distribution plots.

Add Support for numpy>=2.0.0 and Modernize Dependencies
by @guillaume-vignal in #650
- Updates core dependencies to ensure compatibility with NumPy 2.x and future-proof the library.
Update pyproject.toml
by @guerinclement in #651
- Refines build and packaging configuration in line with modern Python standards.
Vectorize Classification Error Computation
by @guillaume-vignal in #666
- Improves performance and scalability when computing classification errors on large datasets.

Synchronize features_dict with Dataset Columns
by @guillaume-vignalin #661
- Ensures consistent alignment between declared features and actual dataset columns.
Rename threading.py to custom_thread.py
by @guillaume-vignal in #667
- Prevents shadowing of Python’s built-in threading module, improving compatibility and reliability.
Change WebApp Favicon
by @guerinclement in #664