def mock_tree_shap_explainer(monkeypatch, request):
"""
Instantiates a TreeShap explainer where both the `_explainer` and `model` atttributes
are mocked
Parameters
----------
request:
Tuple containing number of outputs and model_output string (see TreeShap constructor).
"""
n_outs, model_output = request.param
seed = 0
predictor = MockPredictor(out_dim=n_outs, out_type=model_output, seed=seed)
tree_explainer = MockTreeExplainer(predictor, seed=seed)
monkeypatch.setattr(shap, "TreeExplainer", tree_explainer)
explainer = TreeShap(predictor, model_output=model_output)
return explainer
def make_tree_shap(dirname: Optional[Path] = None) -> TreeShap:
np.random.seed(0)
# get X_train for explainer fit
adult = fetch_adult()
data = adult.data
target = adult.target
data_perm = np.random.permutation(np.c_[data, target])
data = data_perm[:, :-1]
target = data_perm[:, -1]
idx = 30000
X_train, y_train = data[:idx, :], target[:idx]
X_test, y_test = data[idx + 1:, :], target[idx + 1:]
d_train = xgboost.DMatrix(X_train, label=y_train)
d_test = xgboost.DMatrix(X_test, label=y_test)
params = {
"eta": 0.01,
"objective": "binary:logistic",
"subsample": 0.5,
"base_score": np.mean(y_train),
"eval_metric": "logloss",
}
model = xgboost.train(
params,
d_train,
5000,
evals=[(d_test, "test")],
verbose_eval=100,
early_stopping_rounds=20,
)
tree_explainer = TreeShap(model, model_output="raw", task="classification")
tree_explainer.fit(X_train)
if dirname is not None:
tree_explainer.save(dirname)
return tree_explainer
def tree_explainer(rf_classifier, iris_data):
treeshap = TreeShap(predictor=rf_classifier,
model_output='probability',
feature_names=iris_data['metadata']['feature_names'])
treeshap.fit(iris_data['X_train'])
return treeshap