feature importance

IngeoML/analysis.py

@@ -0,0 +1,43 @@
+# Copyright 2024 Mario Graff Guerrero
+
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+
+#     http://www.apache.org/licenses/LICENSE-2.0
+
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from sklearn.metrics import f1_score
+import numpy as np
+
+
+def feature_importance(model, X, y, predictions,
+                       score=None):
+    """Estimate the feature importance of the model"""
+    if score is None:
+        score = lambda y, hy: f1_score(y, hy, average='macro')
+    base = score(y, model.predict(X))
+    hy = np.array([[score(y, j) for j in i]
+                  for i in predictions])
+    return base - hy
+
+
+def predict_shuffle_inputs(model, X, times: int=100):
+    """Predict X by shuffling all the inputs"""
+    X_origin = X.copy()
+    rng = np.random.default_rng()
+    output = []
+    for i in range(X.shape[1]):
+        inner = []
+        for _ in range(times):
+            rng.shuffle(X[:, i])
+            inner.append(model.predict(X))
+        X = X_origin.copy()
+        output.append(np.vstack(inner))
+    return np.array(output)
+
+

IngeoML/tests/test_analysis.py

@@ -0,0 +1,41 @@
+# Copyright 2024 Mario Graff Guerrero
+
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+
+#     http://www.apache.org/licenses/LICENSE-2.0
+
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from sklearn.datasets import load_iris
+from sklearn.metrics import f1_score, make_scorer
+from sklearn.model_selection import ShuffleSplit
+from sklearn.svm import LinearSVC
+from IngeoML.analysis import feature_importance, predict_shuffle_inputs
+
+
+def test_feature_importance():
+    """Test feature importance"""
+
+    X, y = load_iris(return_X_y=True)
+    split = ShuffleSplit(n_splits=1, train_size=0.7).split(X, y)
+    tr, vs = next(split)
+    m = LinearSVC(dual='auto').fit(X[tr], y[tr])
+    predictions = predict_shuffle_inputs(m, X[vs], times=97)
+    diff = feature_importance(m, X[vs], y[vs], predictions)
+    assert diff.shape == (4, 97)
+
+
+def test_predict_shuffle_inputs():
+    """Test predict_shuffle_inputs"""
+
+    X, y = load_iris(return_X_y=True)
+    split = ShuffleSplit(n_splits=1, train_size=0.7).split(X, y)
+    tr, vs = next(split)
+    m = LinearSVC(dual='auto').fit(X[tr], y[tr])
+    hy = predict_shuffle_inputs(m, X[vs])
+    assert hy.shape == (4, 100, vs.shape[0])