def decision_plot(self, X, y):
"""Visualization of the additive feature attribution."""
# Automates single-target slicing
y = super()._check_target_index(y=y)
for index in range(_n_targets(y)):
self.fit(X=X, y=y, index=index)
explainer, shap_values = self.explainer(X=X)
shap.decision_plot(base_value=explainer.expected_value,
shap_values=shap_values,
feature_names=list(X.columns),
show=self.show)
def force_plot(self, X, y):
"""Interactive Javascript visualization of Shapley values."""
shap.initjs()
# Automates single-target slicing
y = super()._check_target_index(y=y)
for index in range(_n_targets(y)):
self.fit(X=X, y=y, index=index)
explainer, shap_values = self.explainer(X=X)
force_plot = display(
shap.force_plot(base_value=explainer.expected_value,
shap_values=shap_values,
features=X,
plot_cmap=['#52A267', '#F0693B'],
feature_names=list(X.columns)))
def summary_plot(self, X, y, plot_type='dot'):
"""Visualizaion of the feature importance and feature effects."""
assert (plot_type in _SHAP_SUMMARY_PLOT_CHOICE)
# Automates single-target slicing
y = super()._check_target_index(y=y)
for index in range(_n_targets(y)):
self.fit(X=X, y=y, index=index)
_, shap_values = self.explainer(X=X)
shap.summary_plot(shap_values=shap_values,
features=X,
plot_type=plot_type,
feature_names=list(X.columns),
show=self.show)
def dependence_plot(self,
X,
y,
interaction_index='auto',
alpha=None,
dot_size=None):
"""Visualization of a feature's effect on a regressor's prediction."""
# Automates single-target slicing
y = super()._check_target_index(y=y)
for index in range(_n_targets(y)):
self.fit(X=X, y=y, index=index)
_, shap_values = self.explainer(X=X)
shap.dependence_plot(ind='rank(0)',
shap_values=shap_values,
features=X,
feature_names=list(X.columns),
cmap=plt.get_cmap('hot'),
interaction_index=interaction_index,
alpha=alpha,
dot_size=dot_size,
show=self.show)