Updates tree decision criterion for scikit-learn 1.3 (#115)

* update for scikit-learn 1.2

* fix compilation issues, updates API

* fix compilation issues

* switch to development version of scikit-learn

* remove normalize, update to the latest API

* Update test_sklearn_stacking.py

* Update test_sklearn_stacking.py

* Update test_piecewise_classifier.py

* update requirements

* fix for scikit-learn==1.3.0

* fix issue

* disable one warning

* lint

* setup

* requirements.txt

* requirements

* remove appveyor

---------

Co-authored-by: xavier dupré <xavier.dupre@gmail.com>

Author: xadupre

.circleci/config.yml

@@ -46,7 +46,7 @@ jobs:
       - run:
           name: Install standard libraries
           command: |
-            python -m pip install scipy matplotlib numpy cython pandas
+            python -m pip install scipy matplotlib numpy cython pandas pyquicksetup
 
       - run:
           name: install dependencies

.gitignore

@@ -80,6 +80,7 @@ build/
 *.pidb
 *.log
 *.scc
+*.so
 *.pyd
 
 # Visual C++ cache files

.local.jenkins.lin.yml

@@ -11,7 +11,6 @@ install:
   - $PYINT -m pip install --upgrade pip
   - $PYINT -m pip install --upgrade --no-cache-dir --no-deps --index http://localhost:8067/simple/ jyquickhelper pyquickhelper cpyquickhelper pandas_streaming --extra-index-url=https://pypi.python.org/simple/
   - $PYINT -m pip install --upgrade --no-cache-dir --no-deps --index http://localhost:8067/simple/ scikit-learn>=0.22 --extra-index-url=https://pypi.python.org/simple/
-  - $PYINT -m pip install -r requirements-win.txt
   - $PYINT -m pip install -r requirements.txt
   - $PYINT --version
   - $PYINT -m pip freeze

_unittests/ut_helpers/test_debug.py

@@ -38,7 +38,7 @@ def test_union_features_reg(self):
                 self.assertNotIn(" object at 0x", text)
                 self.assertIn(") -> (", text)
             else:
-                raise Exception("should not be the case")
+                raise AssertionError("should not be the case")
 
     def test_union_features_cl(self):
         data = numpy.random.randn(4, 5)
@@ -60,7 +60,7 @@ def test_union_features_cl(self):
                 self.assertNotIn(" object at 0x", text)
                 self.assertIn(") -> (", text)
             else:
-                raise Exception("should not be the case")
+                raise AssertionError("should not be the case")
 
 
 if __name__ == "__main__":

_unittests/ut_mlmodel/test_anmf_predictor.py

@@ -45,8 +45,8 @@ def test_anmf_predictor_sparse(self):
         exp = mod.estimator_nmf_.inverse_transform(
             mod.estimator_nmf_.transform(mat))
         got = mod.predict(mat)
-        sc1 = mean_squared_error(mat.todense(), exp)
-        sc2 = mean_squared_error(mat.todense(), got)
+        sc1 = mean_squared_error(numpy.asarray(mat.todense()), exp)
+        sc2 = mean_squared_error(numpy.asarray(mat.todense()), got)
         self.assertGreater(sc1, sc2)
 
         mat2 = numpy.array([[1, 1, 1, 1]], dtype=numpy.float64)
@@ -68,17 +68,17 @@ def test_anmf_predictor_sparse_sparse(self):
         exp = mod.estimator_nmf_.inverse_transform(
             mod.estimator_nmf_.transform(mat))
         got = mod.predict(mat)
-        sc1 = mean_squared_error(mat.todense(), exp)
-        sc2 = mean_squared_error(mat.todense(), got)
+        sc1 = mean_squared_error(numpy.asarray(mat.todense()), exp)
+        sc2 = mean_squared_error(numpy.asarray(mat.todense()), got)
         self.assertGreater(sc1, sc2)
 
         mat2 = numpy.array([[1, 1, 1, 1]], dtype=numpy.float64)
         mat2 = csr_matrix(mat2)
         exp2 = mod.estimator_nmf_.inverse_transform(
             mod.estimator_nmf_.transform(mat2))
         got2 = mod.predict(mat2)
-        sc1 = mean_squared_error(mat2.todense(), exp2)
-        sc2 = mean_squared_error(mat2.todense(), got2)
+        sc1 = mean_squared_error(numpy.asarray(mat2.todense()), exp2)
+        sc2 = mean_squared_error(numpy.asarray(mat2.todense()), got2)
         self.assertGreater(sc1, sc2)
 
     def test_anmf_predictor_positive(self):
@@ -118,8 +118,8 @@ def test_anmf_predictor_positive_sparse(self):
         exp = mod.estimator_nmf_.inverse_transform(
             mod.estimator_nmf_.transform(mat))
         got = mod.predict(mat)
-        sc1 = mean_squared_error(mat.todense(), exp)
-        sc2 = mean_squared_error(mat.todense(), got)
+        sc1 = mean_squared_error(numpy.asarray(mat.todense()), exp)
+        sc2 = mean_squared_error(numpy.asarray(mat.todense()), got)
         self.assertGreater(sc1, sc2)
         mx = numpy.min(got)
         self.assertGreater(mx, 0)

_unittests/ut_mlmodel/test_categories_to_integers.py

@@ -13,7 +13,9 @@
 from pyquickhelper.texthelper import compare_module_version
 from mlinsights.mlmodel import CategoriesToIntegers
 from mlinsights.mlmodel import (
-    test_sklearn_pickle, test_sklearn_clone, test_sklearn_grid_search_cv)
+    run_test_sklearn_pickle,
+    run_test_sklearn_clone,
+    run_test_sklearn_grid_search_cv)
 
 skipped_warnings = (ConvergenceWarning, UserWarning, FitFailedWarning)
 
@@ -103,12 +105,12 @@ def test_categories_to_integers_pickle(self):
         data = os.path.join(os.path.abspath(
             os.path.dirname(__file__)), "data", "adult_set.txt")
         df = pandas.read_csv(data, sep="\t")
-        test_sklearn_pickle(lambda: CategoriesToIntegers(skip_errors=True), df)
+        run_test_sklearn_pickle(lambda: CategoriesToIntegers(skip_errors=True), df)
 
     @ignore_warnings(skipped_warnings)
     def test_categories_to_integers_clone(self):
         self.maxDiff = None
-        test_sklearn_clone(lambda: CategoriesToIntegers())
+        run_test_sklearn_clone(lambda: CategoriesToIntegers())
 
     @ignore_warnings(skipped_warnings)
     def test_categories_to_integers_grid_search(self):
@@ -119,19 +121,19 @@ def test_categories_to_integers_grid_search(self):
         y = df['income']  # pylint: disable=E1136
         pipe = make_pipeline(CategoriesToIntegers(),
                              LogisticRegression())
-        self.assertRaise(lambda: test_sklearn_grid_search_cv(
+        self.assertRaise(lambda: run_test_sklearn_grid_search_cv(
             lambda: pipe, df), ValueError)
         if (compare_module_version(sklver, "0.24") >= 0 and  # pylint: disable=R1716
                 compare_module_version(pandas.__version__, "1.3") < 0):
             self.assertRaise(
-                lambda: test_sklearn_grid_search_cv(
+                lambda: run_test_sklearn_grid_search_cv(
                     lambda: pipe, X, y, categoriestointegers__single=[True, False]),
                 ValueError, "Unable to find category value")
         pipe = make_pipeline(CategoriesToIntegers(),
                              Imputer(strategy='most_frequent'),
                              LogisticRegression(n_jobs=1))
         try:
-            res = test_sklearn_grid_search_cv(
+            res = run_test_sklearn_grid_search_cv(
                 lambda: pipe, X, y, categoriestointegers__single=[True, False],
                 categoriestointegers__skip_errors=[True])
         except AttributeError as e:

_unittests/ut_mlmodel/test_classification_kmeans.py

@@ -15,8 +15,8 @@
 from pyquickhelper.pycode import ExtTestCase
 from pyquickhelper.texthelper import compare_module_version
 from mlinsights.mlmodel import (
-    ClassifierAfterKMeans, test_sklearn_pickle, test_sklearn_clone,
-    test_sklearn_grid_search_cv)
+    ClassifierAfterKMeans, run_test_sklearn_pickle,
+    run_test_sklearn_clone, run_test_sklearn_grid_search_cv)
 
 
 class TestClassifierAfterKMeans(ExtTestCase):
@@ -58,24 +58,24 @@ def test_classification_kmeans_pickle(self):
         iris = datasets.load_iris()
         X, y = iris.data, iris.target
         try:
-            test_sklearn_pickle(lambda: ClassifierAfterKMeans(), X, y)
+            run_test_sklearn_pickle(lambda: ClassifierAfterKMeans(), X, y)
         except AttributeError as e:
             if compare_module_version(sklver, "0.24") < 0:
                 return
             raise e
 
     def test_classification_kmeans_clone(self):
         self.maxDiff = None
-        test_sklearn_clone(lambda: ClassifierAfterKMeans())
+        run_test_sklearn_clone(lambda: ClassifierAfterKMeans())
 
     @ignore_warnings(category=ConvergenceWarning)
     def test_classification_kmeans_grid_search(self):
         iris = datasets.load_iris()
         X, y = iris.data, iris.target
-        self.assertRaise(lambda: test_sklearn_grid_search_cv(
+        self.assertRaise(lambda: run_test_sklearn_grid_search_cv(
             lambda: ClassifierAfterKMeans(), X, y), ValueError)
         try:
-            res = test_sklearn_grid_search_cv(
+            res = run_test_sklearn_grid_search_cv(
                 lambda: ClassifierAfterKMeans(),
                 X, y, c_n_clusters=[2, 3])
         except AttributeError as e:

_unittests/ut_mlmodel/test_decision_tree_logistic_regression.py

@@ -12,9 +12,8 @@
 from sklearn.tree import DecisionTreeClassifier
 from pyquickhelper.pycode import ExtTestCase
 from mlinsights.mlmodel import (
-    test_sklearn_pickle, test_sklearn_clone, test_sklearn_grid_search_cv,
-    DecisionTreeLogisticRegression
-)
+    run_test_sklearn_pickle, run_test_sklearn_clone,
+    run_test_sklearn_grid_search_cv, DecisionTreeLogisticRegression)
 from mlinsights.mltree import predict_leaves
 
 
@@ -63,22 +62,23 @@ def test_classifier_pickle(self):
         X = random(100)
         Y = X > 0.5  # pylint: disable=W0143
         X = X.reshape((100, 1))  # pylint: disable=E1101
-        test_sklearn_pickle(lambda: LogisticRegression(), X, Y)
-        test_sklearn_pickle(lambda: DecisionTreeLogisticRegression(
+        run_test_sklearn_pickle(lambda: LogisticRegression(), X, Y)
+        run_test_sklearn_pickle(lambda: DecisionTreeLogisticRegression(
             fit_improve_algo=None), X, Y)
 
     def test_classifier_clone(self):
-        test_sklearn_clone(
+        run_test_sklearn_clone(
             lambda: DecisionTreeLogisticRegression(fit_improve_algo=None))
 
     def test_classifier_grid_search(self):
         X = random(100)
         Y = X > 0.5  # pylint: disable=W0143
         X = X.reshape((100, 1))  # pylint: disable=E1101
-        self.assertRaise(lambda: test_sklearn_grid_search_cv(
+        self.assertRaise(lambda: run_test_sklearn_grid_search_cv(
             lambda: DecisionTreeLogisticRegression(fit_improve_algo=None), X, Y), ValueError)
-        res = test_sklearn_grid_search_cv(lambda: DecisionTreeLogisticRegression(fit_improve_algo=None),
-                                          X, Y, max_depth=[2, 3])
+        res = run_test_sklearn_grid_search_cv(
+            lambda: DecisionTreeLogisticRegression(fit_improve_algo=None),
+            X, Y, max_depth=[2, 3])
         self.assertIn('model', res)
         self.assertIn('score', res)
         self.assertGreater(res['score'], 0)