def get_interaction_values(self, n_samples_max=None, selection=None):
"""
Compute shap interaction values for each row of x_init.
This function is only available for explainer of type TreeExplainer (used for tree based models).
Please refer to the official tree shap paper for more information : https://arxiv.org/pdf/1802.03888.pdf
Parameters
----------
n_samples_max : int, optional
Limit the number of points for which we compute the interactions.
selection : list, optional
Contains list of index, subset of the input DataFrame that we want to plot
Returns
-------
np.ndarray
Shap interaction values for each sample as an array of shape (# samples x # features x # features).
"""
x = copy.deepcopy(self.x_init)
if selection:
x = x.loc[selection]
if hasattr(self, 'x_interaction'):
if self.x_interaction.equals(x[:n_samples_max]):
return self.interaction_values
self.x_interaction = x[:n_samples_max]
self.interaction_values = get_shap_interaction_values(self.x_interaction, self.explainer)
return self.interaction_values
def test_shap_interaction_values_2(self):
"""
Unit test shap_interaction_values function
"""
for model in [
ske.RandomForestRegressor(n_estimators=1),
ske.RandomForestClassifier(n_estimators=1)
]:
self.x_df = self.x_df.astype(float)
model.fit(self.x_df, self.y_df)
print(model)
explainer = shap.TreeExplainer(model)
interaction_values = get_shap_interaction_values(
self.x_df, explainer)
assert interaction_values.shape[0] == self.x_df.shape[0]
assert interaction_values.shape[1] == self.x_df.shape[1]
assert interaction_values.shape[2] == self.x_df.shape[1]