Updated partial dependence methods to support non-numeric columns

evalml/model_understanding/graphs.py

@@ -20,6 +20,7 @@
 from evalml.objectives.utils import get_objective
 from evalml.problem_types import ProblemTypes
 from evalml.utils import import_or_raise, jupyter_check
+from evalml.utils.gen_utils import numeric_dtypes
 
 
 def confusion_matrix(y_true, y_predicted, normalize_method='true'):
@@ -406,7 +407,7 @@ def partial_dependence(pipeline, X, feature, grid_resolution=100):
 
     Arguments:
         pipeline (PipelineBase or subclass): Fitted pipeline
-        X (pd.DataFrame, npermutation importance.array): The input data used to generate a grid of values
+        X (pd.DataFrame, np.array): The input data used to generate a grid of values
             for feature where partial dependence will be calculated at
         feature (int, string): The target features for which to create the partial dependence plot for.
             If feature is an int, it must be the index of the feature to use.
@@ -417,6 +418,10 @@ def partial_dependence(pipeline, X, feature, grid_resolution=100):
             over all samples of X and the values used to calculate those predictions.
 
     """
+    if not isinstance(X, pd.DataFrame):
+        X = pd.DataFrame(X)
+    if X[feature].dtype not in numeric_dtypes:
+        raise ValueError(f"Partial dependence is is currently only supported for numeric dtypes.")
     if pipeline.model_family == ModelFamily.BASELINE:
         raise ValueError("Partial dependence plots are not supported for Baseline pipelines")
     if not pipeline._is_fitted:
@@ -445,7 +450,7 @@ def graph_partial_dependence(pipeline, X, feature, grid_resolution=100):
 
     Arguments:
         pipeline (PipelineBase or subclass): Fitted pipeline
-        X (pd.DataFrame, npermutation importance.array): The input data used to generate a grid of values
+        X (pd.DataFrame, np.array): The input data used to generate a grid of values
             for feature where partial dependence will be calculated at
         feature (int, string): The target feature for which to create the partial dependence plot for.
             If feature is an int, it must be the index of the feature to use.

evalml/tests/model_understanding_tests/test_graphs.py

@@ -638,6 +638,24 @@ def test_partial_dependence_string_feature_name(logistic_regression_binary_pipel
     assert not part_dep.isnull().any(axis=None)
 
 
+def test_partial_dependence_nonsupported_dtypes(logistic_regression_binary_pipeline_class):
+    X = pd.DataFrame({
+        "categorical col": pd.Series(["zero", "one", "two", "zero", "three"], dtype='category'),
+        "object col": ["b", "b", "a", "c", "d"],
+        "bool col": [True, False, False, True, True],
+        "int col": [0, 1, 2, 4, 3]
+    })
+    y = pd.Series([1, 1, 0, 0, 1])
+    pipeline = logistic_regression_binary_pipeline_class(parameters={})
+    pipeline.fit(X, y)
+    with pytest.raises(ValueError, match="Partial dependence is is currently only supported for numeric dtypes"):
+        partial_dependence(pipeline, X, feature="categorical col", grid_resolution=20)
+    with pytest.raises(ValueError, match="Partial dependence is is currently only supported for numeric dtypes"):
+        partial_dependence(pipeline, X, feature="object col", grid_resolution=20)
+    with pytest.raises(ValueError, match="Partial dependence is is currently only supported for numeric dtypes"):
+        partial_dependence(pipeline, X, feature="bool col", grid_resolution=20)
+
+
 def test_partial_dependence_with_non_numeric_columns(linear_regression_pipeline_class):
     X = pd.DataFrame({'numeric': [1, 2, 3, 0], 'also numeric': [2, 3, 4, 1], 'string': ['a', 'b', 'a', 'c'], 'also string': ['c', 'b', 'a', 'd']})
     y = [0, 0.2, 1.4, 1]