Explainerdashboard

https://github.com/oegedijk/explainerdashboard

https://explainerdashboard.readthedocs.io/en/latest/

Installation

conda create -n test_env python=3.8

conda activate test_env

pip install explainerdashboard

Windows

> choco install graphviz

Quick start

https://colab.research.google.com/github/oegedijk/explainerdashboard/blob/master/explainer_examples.ipynb

from explainerdashboard.datasets import titanic_fare, titanic_names, feature_descriptions

from sklearn.ensemble import RandomForestRegressor

from explainerdashboard import RegressionExplainer

X_train, y_train, X_test, y_test = titanic_fare()

model = RandomForestRegressor(n_estimators=50, max_depth=5)

model.fit(X_train, y_train)

train_names, test_names = titanic_names()

explainer = RegressionExplainer(model, X_test, y_test,

cats=['Sex', 'Deck', 'Embarked'],

idxs=test_names,

target='Fare',

descriptions=feature_descriptions,

units="$")

Useage

model = RandomForestRegressor(n_jobs=-1, random_state=42)

model = model.fit(X_train, y_train)

explainer = RegressionExplainer(

model,

X_test,

y_test,

cats=["COL1", "COL2", "COL3"],

descriptions={"COL1": "aaa", "COL2", "bbb", "COL2", "ccc"},

target='Nice Name', # defaults to y.name

units='$',

# higher_is_better=False,

shap="tree",

)

Feature importance

explainer.mean_abs_shap_df(cats=True, cutoff=0.01)

explainer.permutation_importances_df(cats=False, topx=12)

Contributions

index = 0 # explain prediction for first row of X_test

explainer.contrib_df(index, cats=True, topx=8)

Hide components

https://github.com/oegedijk/explainerdashboard#hiding-components

Dashboard

ExplainerDashboard(

explainer,

title='TITLE',

orientation='horizontal',

hide_orientation=True,

# Tabs

importances=True,

model_summary=False,

contributions=True,

whatif=True,

shap_dependence=True,

shap_interaction=False

decision_trees=True

).run()

Plots

explainer.plot_importances(kind='shap', topx=5, cats=True)

explainer.plot_shap_summary(topx=10, cats=True)

explainer.plot_shap_contributions(index, cats=True, topx=8) # sort='high-to-low', orientation='horizontal'

explainer.plot_shap_contributions(name, cats=False)

explainer.plot_shap_dependence("Age") # color_col="Sex", highlight_index=5