Failed to guess the type of shap explainer to use

[raybellwaves]

First time using this code. Thanks.

When I try

explainer = RegressionExplainer(model, X_test, y_test)

I get

ValueError: Failed to guess the type of shap explainer to use. Please explicitly pass either shap='tree', 'linear', deep' or 'kernel'.

i'm not sure where to put this

[oegedijk]

Ah, yeah, I could clarify that error message. In an earlier iteration of explainerdashboard you had to pass a parameter on what kind of shap explainer your model needed (shap.TreeExplainer, shap.LinearExplainer, shap.KernelExplainer, etc), by passing a parameter to the explainer e.g. shap='tree'.

Then I made it default to shap='guess' where the explainer tries to guess what kind of shap explainer is needed from the name of the model class. In case it fails to do this you get the above message.

What is the model that you fitted?

If it's a linear model you would pass it for example as:

explainer = RegressionExplainer(model, X_test, y_test, shap='linear')

[raybellwaves]

Thanks. It's a sklearn.ensemble.RandomForestRegressor

so i'll use

explainer = RegressionExplainer(model, X_test, y_test, shap='tree')

[oegedijk]

That's weird, that should definitely be detected (and it works in the tests that use a RandomForestRegressor as well).

What do you see when your print(str(type(model)))?

For reference, these models should all automatically get shap='tree' and shap='linear':

tree_models = ['RandomForestClassifier', 'RandomForestRegressor',
'DecisionTreeClassifier', 'DecisionTreeRegressor',
'ExtraTreesClassifier', 'ExtraTreesRegressor',
'GradientBoostingClassifier', 'GradientBoostingRegressor',
'HistGradientBoostingClassifier', 'HistGradientBoostingRegressor',
'XGBClassifier', 'XGBRegressor',
'LGBMClassifier', 'LGBMRegressor',
'CatBoostClassifier', 'CatBoostRegressor',
'NGClassifier', 'NGBRegressor',
'GBTClassifier', ' GBTRegressor',
'IsolationForest'
]
linear_models = ['LinearRegression', 'LogisticRegression',
'Ridge', 'Lasso', 'ElasticNet']

[raybellwaves]

Actually this follows #11

Probably because it was a pipeline it's wasn't familiar with the type of model.

[oegedijk]

Ah, got it. Will add supporting sklearn pipelines to the roadmap!