SHAP is a useful model interpretation tool that provides a variety of visualization methods. However, most of the plots in SHAP are drawn with matplotlib, which is neither interactive nor easy to modify.
This project reimplements these images with Poltly, in which way model interpretation diagrams can be embedded in web pages as interactive, customizable components.
- waterfall
- local dependence
- summary (beeswarm)
- decision
- dependence (scatter)
Not all drawing details have been implemented
pip install -r requirements_dev.txt
pytest test/testall.py
