Shapley Values with H2O AutoML Example (ML Interpretability)

Example using H2O's AutoML with the SHAP (SHapley Additive exPlanations) library

Task List

• Classification Example
• Regression Example
• Use Plotly for visualizations

Notebooks

Filename	Description
shap_h2o_automl_classification.ipynb	jupyter notebook w/ classification examples
shap_h2o_automl_classification.html	html version for viewing force_plots
plotly_force_plot.ipynb	jupyter notebook plotly force-plots
plotly_force_plot.html	html version plotly force-plots

Step-by-step guide

[1]

conda env create -f environment.yml

[2]

source activate shap_h2o

[3] open jupyter notebook

References:

• An Introduction to Machine Learning Interpretability
• Interpretable Machine Learning
• A Unified Approach to Interpreting Model Predictions
• SHAP (SHapley Additive exPlanations)
• https://shap.readthedocs.io/en/latest/