Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Displaying the top_k features with largest shap value magnitudes #1374

Merged
merged 7 commits into from
Nov 6, 2020
Merged

Displaying the top_k features with largest shap value magnitudes #1374

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
46dd988
Displaying the top_k features with largest shap value magnitudes
freddyaboulton Oct 29, 2020
b020e35
Adding PR 1360 to release notes.
freddyaboulton Oct 30, 2020
9254bb5
Updating broken tests in test_explainers to reflect top_k rows by mag…
freddyaboulton Oct 30, 2020
e20008b
Updating comment.
freddyaboulton Oct 30, 2020
6796e99
Removing linebreak in release notes for pr 1374.
freddyaboulton Oct 30, 2020
f3b6242
Sorting shap values in _make_rows via lambda function.
freddyaboulton Nov 4, 2020
2afa992
Not changing notebook python version in model understanding user guide.
freddyaboulton Nov 4, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions docs/source/release_notes.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@ Release Notes
* Enhancements
* Added ability to freeze hyperparameters for AutoMLSearch :pr:`1284`
* Added the index id to the ``explain_predictions_best_worst`` output to help users identify which rows in their data are included :pr:`1365`
* The top_k features displayed in ``explain_predictions_*`` functions are now determined by the magnitude of shap values as opposed to the ``top_k`` largest and smallest shap values. :pr:`1374`
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

👍

* Fixes
* Updated enum classes to show possible enum values as attributes :pr:`1391`
* Changes
Expand All @@ -17,6 +18,12 @@ Release Notes
* Removed ``category_encoders`` from test-requirements.txt :pr:`1373`
* Tweak codecov.io settings again to avoid flakes :pr:`1413`

.. warning::

**Breaking Changes**
* The ``top_k`` and ``top_k_features`` parameters in ``explain_predictions_*`` functions now return ``k`` features as opposed to ``2 * k`` features :pr:`1374`
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

👍



**v0.15.0 Oct. 29, 2020**
* Enhancements
* Added stacked ensemble component classes (``StackedEnsembleClassifier``, ``StackedEnsembleRegressor``) :pr:`1134`
Expand Down
7 changes: 4 additions & 3 deletions docs/source/user_guide/model_understanding.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -330,7 +330,7 @@
"from evalml.model_understanding.prediction_explanations import explain_prediction\n",
"\n",
"table = explain_prediction(pipeline=pipeline, input_features=X.iloc[3:4],\n",
" training_data=X, include_shap_values=True)\n",
" training_data=X, top_k=6, include_shap_values=True)\n",
"print(table)"
]
},
Expand Down Expand Up @@ -367,7 +367,7 @@
"from evalml.model_understanding.prediction_explanations import explain_predictions_best_worst\n",
"\n",
"report = explain_predictions_best_worst(pipeline=pipeline, input_features=X, y_true=y,\n",
" include_shap_values=True, num_to_explain=2)\n",
" include_shap_values=True, top_k_features=6, num_to_explain=2)\n",
"\n",
"print(report)"
]
Expand Down Expand Up @@ -431,7 +431,8 @@
"source": [
"import json\n",
"report = explain_predictions_best_worst(pipeline=pipeline, input_features=X, y_true=y,\n",
" num_to_explain=1, include_shap_values=True, output_format=\"dict\")\n",
" num_to_explain=1, top_k_features=6,\n",
" include_shap_values=True, output_format=\"dict\")\n",
"print(json.dumps(report, indent=2))"
]
}
Expand Down
11 changes: 6 additions & 5 deletions evalml/model_understanding/prediction_explanations/_user_interface.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,12 +26,13 @@ def _make_rows(shap_values, normalized_values, pipeline_features, top_k, include
list(str)
"""
tuples = [(value[0], feature_name) for feature_name, value in normalized_values.items()]
tuples = sorted(tuples)

if len(tuples) <= 2 * top_k:
features_to_display = reversed(tuples)
else:
features_to_display = tuples[-top_k:][::-1] + tuples[:top_k][::-1]
# Sort the features s.t the top_k w the largest shap value magnitudes are the first
# top_k elements
tuples = sorted(tuples, key=lambda x: abs(x[0]), reverse=True)

# Then sort such that the SHAP values go from most positive to most negative
features_to_display = reversed(sorted(tuples[:top_k]))

rows = []
for value, feature_name in features_to_display:
Expand Down
76 changes: 34 additions & 42 deletions evalml/tests/model_understanding_tests/prediction_explanations_tests/test_explainers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,27 +34,25 @@ def test_explain_prediction_value_error(test_features):

explain_prediction_answer = """Feature Name Feature Value Contribution to Prediction
=========================================================
d 40.00 ++++
a 10.00 +++
c 30.00 --
b 20.00 ----""".splitlines()
d 40.00 +++++
b 20.00 -----""".splitlines()

explain_prediction_regression_dict_answer = {
"explanations": [{
"feature_names": ["d", "a", "c", "b"],
"feature_values": [40, 10, 30, 20],
"qualitative_explanation": ["++++", "+++", "--", "----"],
"quantitative_explanation": [None, None, None, None],
"feature_names": ["d", "b"],
"feature_values": [40, 20],
"qualitative_explanation": ["+++++", "-----"],
"quantitative_explanation": [None, None],
"class_name": None
}]
}

explain_prediction_binary_dict_answer = {
"explanations": [{
"feature_names": ["d", "a", "c", "b"],
"feature_values": [40, 10, 30, 20],
"qualitative_explanation": ["++++", "+++", "--", "----"],
"quantitative_explanation": [None, None, None, None],
"feature_names": ["d", "b"],
"feature_values": [40, 20],
"qualitative_explanation": ["+++++", "-----"],
"quantitative_explanation": [None, None],
"class_name": "class_1"
}]
}
Expand All @@ -63,10 +61,8 @@ def test_explain_prediction_value_error(test_features):

Feature Name Feature Value Contribution to Prediction
=========================================================
a 10.00 +
b 20.00 +
c 30.00 -
d 40.00 -
a 10.00 +++++
c 30.00 ---


Class: class_1
Expand All @@ -75,36 +71,32 @@ def test_explain_prediction_value_error(test_features):
=========================================================
a 10.00 +++
b 20.00 ++
c 30.00 -
d 40.00 --


Class: class_2

Feature Name Feature Value Contribution to Prediction
=========================================================
a 10.00 +
b 20.00 +
c 30.00 ---
d 40.00 ---
""".splitlines()

explain_prediction_multiclass_dict_answer = {
"explanations": [
{"feature_names": ["a", "b", "c", "d"],
"feature_values": [10, 20, 30, 40],
"qualitative_explanation": ["+", "+", "-", "-"],
"quantitative_explanation": [None] * 4,
{"feature_names": ["a", "c"],
"feature_values": [10, 30],
"qualitative_explanation": ["+++++", "---"],
"quantitative_explanation": [None, None],
"class_name": "class_0"},
{"feature_names": ["a", "b", "c", "d"],
"feature_values": [10, 20, 30, 40],
"qualitative_explanation": ["+++", "++", "-", "--"],
"quantitative_explanation": [None] * 4,
{"feature_names": ["a", "b"],
"feature_values": [10, 20],
"qualitative_explanation": ["+++", "++"],
"quantitative_explanation": [None, None],
"class_name": "class_1"},
{"feature_names": ["a", "b", "c", "d"],
"feature_values": [10, 20, 30, 40],
"qualitative_explanation": ["+", "+", "---", "---"],
"quantitative_explanation": [None] * 4,
{"feature_names": ["c", "d"],
"feature_values": [30, 40],
"qualitative_explanation": ["---", "---"],
"quantitative_explanation": [None, None],
"class_name": "class_2"},
]
}
Expand All @@ -113,36 +105,36 @@ def test_explain_prediction_value_error(test_features):
@pytest.mark.parametrize("problem_type,output_format,shap_values,normalized_shap_values,answer",
[(ProblemTypes.REGRESSION,
"text",
{"a": [1], "b": [-2], "c": [-0.25], "d": [2]},
{"a": [0.5], "b": [-0.75], "c": [-0.25], "d": [0.75]},
{"a": [1], "b": [-2.1], "c": [-0.25], "d": [2.3]},
{"a": [0.5], "b": [-2.1], "c": [-0.25], "d": [2.3]},
explain_prediction_answer),
(ProblemTypes.REGRESSION,
"dict",
{"a": [1], "b": [-2], "c": [-0.25], "d": [2]},
{"a": [0.5], "b": [-0.75], "c": [-0.25], "d": [0.75]},
{"a": [1], "b": [-2.1], "c": [-0.25], "d": [2.3]},
{"a": [0.5], "b": [-2.1], "c": [-0.25], "d": [2.3]},
explain_prediction_regression_dict_answer
),
(ProblemTypes.BINARY,
"text",
[{}, {"a": [1], "b": [-2], "c": [-0.25], "d": [2]}],
[{}, {"a": [0.5], "b": [-0.75], "c": [-0.25], "d": [0.75]}],
[{}, {"a": [0.5], "b": [-0.89], "c": [0.33], "d": [0.89]}],
[{}, {"a": [0.5], "b": [-0.89], "c": [-0.25], "d": [0.89]}],
explain_prediction_answer),
(ProblemTypes.BINARY,
"dict",
[{}, {"a": [1], "b": [-2], "c": [-0.25], "d": [2]}],
[{}, {"a": [0.5], "b": [-0.75], "c": [-0.25], "d": [0.75]}],
[{}, {"a": [0.5], "b": [-0.89], "c": [0.33], "d": [0.89]}],
[{}, {"a": [0.5], "b": [-0.89], "c": [-0.25], "d": [0.89]}],
explain_prediction_binary_dict_answer),
(ProblemTypes.MULTICLASS,
"text",
[{}, {}, {}],
[{"a": [0.1], "b": [0.09], "c": [-0.04], "d": [-0.06]},
[{"a": [1.1], "b": [0.09], "c": [-0.53], "d": [-0.06]},
{"a": [0.53], "b": [0.24], "c": [-0.15], "d": [-0.22]},
{"a": [0.03], "b": [0.02], "c": [-0.42], "d": [-0.47]}],
explain_prediction_multiclass_answer),
(ProblemTypes.MULTICLASS,
"dict",
[{}, {}, {}],
[{"a": [0.1], "b": [0.09], "c": [-0.04], "d": [-0.06]},
[{"a": [1.1], "b": [0.09], "c": [-0.53], "d": [-0.06]},
{"a": [0.53], "b": [0.24], "c": [-0.15], "d": [-0.22]},
{"a": [0.03], "b": [0.02], "c": [-0.42], "d": [-0.47]}],
explain_prediction_multiclass_dict_answer)
Expand Down
Loading