search

MAIF / shapash

🔅 Shapash: User-friendly Explainability and Interpretability to Develop Reliable and Transparent Machine Learning Models
https://maif.github.io/shapash/
Apache License 2.0
2.72k stars 331 forks source link

NB and shapash problem #257

Open npapan69 opened 2 years ago

npapan69 commented 2 years ago

Dear All,

I have an error message when trying to compile xpl if I use a naive bayes model

model not supported by shapash, please compute contributions by yourself before using shapash

Any help with this will be appreciated

many thanks in advance

ThomasBouche commented 2 years ago

Hi,

We don't compute contributions with naive bayes model, but you can compute your own contributions (with shap.KernelExplainer for example). Then you can compute Shapash with your contributions. You can see this tutorial : https://github.com/MAIF/shapash/blob/master/tutorial/explainer/tuto-expl01-Shapash-Viz-using-Shap-contributions.ipynb

npapan69 commented 2 years ago

Dear Thomas, Many thanks for your response. Is it possible to define a probability threshold other than 0.5 on the predictor?