Fixes #95, PiecewiseRegressor, makes sure target are vectors

.travis.yml

@@ -7,7 +7,7 @@ matrix:
   - python: 3.9
     name: "Py39-024+"
     env:
-      - sklver=">=0.24.0"
+      - sklver=">=0.24.2"
       - jlver=">=1.0"
   - python: 3.8
     name: "Py38-023"

_doc/sphinxdoc/source/conf.py

@@ -5,7 +5,7 @@
 import sys
 import os
 import alabaster
-from pyquickhelper.helpgen.default_conf import set_sphinx_variables, get_default_stylesheet
+from pyquickhelper.helpgen.default_conf import set_sphinx_variables
 
 
 sys.path.insert(0, os.path.abspath(os.path.join(os.path.split(__file__)[0])))
@@ -20,8 +20,9 @@
                      title="mlinsights", book=True)
 
 blog_root = "http://www.xavierdupre.fr/app/mlinsights/helpsphinx/"
-html_context = {'css_files': get_default_stylesheet() +
-                ['_static/my-styles.css']}
+
+html_css_files = ['my-styles.css']
+
 html_logo = "phdoc_static/project_ico.png"
 html_sidebars = {}
 language = "en"

_unittests/ut_mlmodel/test_categories_to_integers.py

@@ -119,10 +119,11 @@ def test_categories_to_integers_grid_search(self):
                              LogisticRegression())
         self.assertRaise(lambda: test_sklearn_grid_search_cv(
             lambda: pipe, df), ValueError)
-        self.assertRaise(
-            lambda: test_sklearn_grid_search_cv(
-                lambda: pipe, X, y, categoriestointegers__single=[True, False]),
-            ValueError, "Unable to find category value")
+        if compare_module_version(sklver, "0.24") >= 0:
+            self.assertRaise(
+                lambda: 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))

_unittests/ut_mlmodel/test_classification_kmeans.py

@@ -57,7 +57,12 @@ def test_classification_kmeans_intercept_weights(self):
     def test_classification_kmeans_pickle(self):
         iris = datasets.load_iris()
         X, y = iris.data, iris.target
-        test_sklearn_pickle(lambda: ClassifierAfterKMeans(), X, y)
+        try:
+            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

_unittests/ut_mlmodel/test_piecewise_regressor.py

@@ -7,6 +7,8 @@
 from numpy.random import random
 import pandas
 from sklearn.linear_model import LinearRegression
+from sklearn.datasets import make_regression
+from sklearn.tree import DecisionTreeRegressor
 from pyquickhelper.pycode import ExtTestCase, ignore_warnings
 from mlinsights.mlmodel import test_sklearn_pickle, test_sklearn_clone, test_sklearn_grid_search_cv
 from mlinsights.mlmodel.piecewise_estimator import PiecewiseRegressor
@@ -148,6 +150,23 @@ def test_piecewise_regressor_grid_search(self):
         self.assertGreater(res['score'], 0)
         self.assertLesser(res['score'], 1)
 
+    def test_piecewise_regressor_issue(self):
+        X, y = make_regression(10000, n_features=1, n_informative=1,  # pylint: disable=W0632
+                               n_targets=1)
+        y = y.reshape((-1, 1))
+        model = PiecewiseRegressor(
+            binner=DecisionTreeRegressor(min_samples_leaf=300))
+        model.fit(X, y)
+        vvc = model.predict(X)
+        self.assertEqual(vvc.shape, (X.shape[0], ))
+
+    def test_piecewise_regressor_raise(self):
+        X, y = make_regression(10000, n_features=2, n_informative=2,  # pylint: disable=W0632
+                               n_targets=2)
+        model = PiecewiseRegressor(
+            binner=DecisionTreeRegressor(min_samples_leaf=300))
+        self.assertRaise(lambda: model.fit(X, y), RuntimeError)
+
 
 if __name__ == "__main__":
     unittest.main()

mlinsights/mlmodel/_kmeans_022.py

@@ -9,7 +9,7 @@
 from sklearn.utils.sparsefuncs_fast import assign_rows_csr  # pylint: disable=W0611,E0611
 try:
     from sklearn.cluster._kmeans import _check_sample_weight
-except ImportError:
+except ImportError:  # pragma: no cover
     from sklearn.cluster._kmeans import (
         _check_normalize_sample_weight as _check_sample_weight)
 from sklearn.metrics.pairwise import pairwise_distances_argmin_min

mlinsights/mlmodel/_kmeans_constraint_.py

@@ -234,7 +234,8 @@ def _compute_strategy_coefficient(distances, strategy, labels):
         ar = numpy.arange(distances.shape[0])
         dist = distances[ar, labels]
         return distances - dist[:, numpy.newaxis]
-    raise ValueError("Unknwon strategy '{0}'.".format(strategy))
+    raise ValueError(  # pragma: no cover
+        "Unknwon strategy '{0}'.".format(strategy))
 
 
 def _randomize_index(index, weights):

mlinsights/mlmodel/decision_tree_logreg.py

@@ -422,7 +422,7 @@ def _fit_parallel(self, X, y, sample_weight):
 
     def _fit_perpendicular(self, X, y, sample_weight):
         "Implements the perpendicular strategy."
-        raise NotImplementedError()
+        raise NotImplementedError()  # pragma: no cover
 
     def predict(self, X):
         """

mlinsights/mlmodel/extended_features.py

@@ -71,7 +71,7 @@ def _get_feature_names_poly(self, input_features=None):
             input_features = ["x%d" %
                               i for i in range(0, self.n_input_features_)]
         elif len(input_features) != self.n_input_features_:
-            raise ValueError(
+            raise ValueError(  # pragma: no cover
                 "input_features should contain {} strings.".format(
                     self.n_input_features_))
 
@@ -125,7 +125,7 @@ def fit(self, X, y=None):
             return self._fit_poly(X, y)
         elif self.kind == 'poly-slow':
             return self._fit_poly(X, y)
-        raise ValueError(
+        raise ValueError(  # pragma: no cover
             "Unknown extended features '{}'.".format(self.kind))
 
     def _fit_poly(self, X, y=None):
@@ -148,7 +148,8 @@ def transform(self, X):
         """
         n_features = X.shape[1]
         if n_features != self.n_input_features_:
-            raise ValueError("X shape does not match training shape")
+            raise ValueError(  # pragma: no cover
+                "X shape does not match training shape")
         if self.kind == 'poly':
             return self._transform_poly(X)
         if self.kind == 'poly-slow':

mlinsights/mlmodel/piecewise_estimator.py

@@ -13,7 +13,7 @@
 from sklearn.utils.fixes import _joblib_parallel_args
 try:
     from tqdm import tqdm
-except ImportError:
+except ImportError:  # pragma: no cover
     pass
 
 
@@ -223,6 +223,12 @@ def fit(self, X, y, sample_weight=None):
         * `dim_`: dimension of the output
         * `mean_`: average targets
         """
+        if len(y.shape) == 2:
+            if y.shape[-1] == 1:
+                y = y.ravel()
+            else:
+                raise RuntimeError(
+                    "This regressor only works with single dimension targets.")
         if isinstance(X, pandas.DataFrame):
             X = X.values
         if isinstance(X, list):

mlinsights/mlmodel/sklearn_testing.py

@@ -259,6 +259,8 @@ def test_sklearn_grid_search_cv(fct_model, X, y=None, sample_weight=None, **grid
     clf = GridSearchCV(pipe, parameters)
     if y_train is None and w_train is None:
         clf.fit(X_train)
+    elif w_train is None:
+        clf.fit(X_train, y_train)  # pylint: disable=E1121
     else:
         clf.fit(X_train, y_train, w_train)  # pylint: disable=E1121
     score = clf.score(X_test, y_test)

mlinsights/mlmodel/sklearn_transform_inv_fct.py

@@ -46,12 +46,13 @@ def __init__(self, fct, fct_inv=None):
         BaseReciprocalTransformer.__init__(self)
         if isinstance(fct, str):
             if fct_inv is not None:
-                raise ValueError(
+                raise ValueError(  # pragma: no cover
                     "If fct is a function name, fct_inv must not be specified.")
             opts = self.__class__.available_fcts()
             if fct not in opts:
-                raise ValueError("Unknown fct '{}', it should in {}.".format(
-                    fct, list(sorted(opts))))
+                raise ValueError(  # pragma: no cover
+                    "Unknown fct '{}', it should in {}.".format(
+                        fct, list(sorted(opts))))
         else:
             if fct_inv is None:
                 raise ValueError(
@@ -117,7 +118,8 @@ def fit(self, X=None, y=None, sample_weight=None):
         Defines a random permutation over the targets.
         """
         if y is None:
-            raise RuntimeError("targets cannot be empty.")
+            raise RuntimeError(  # pragma: no cover
+                "targets cannot be empty.")
         num = numpy.issubdtype(y.dtype, numpy.floating)
         perm = {}
         for u in y.ravel():
@@ -141,7 +143,7 @@ def fit(self, X=None, y=None, sample_weight=None):
 
     def _check_is_fitted(self):
         if not hasattr(self, 'permutation_'):
-            raise NotFittedError(
+            raise NotFittedError(  # pragma: no cover
                 "This instance {} is not fitted yet. Call 'fit' with "
                 "appropriate arguments before using this method.".format(
                     type(self)))
@@ -169,8 +171,9 @@ def _find_closest(self, cl):
             return float(res)
         if self.knn_perm_.dtype in (numpy.int32, numpy.int64):
             return int(res)
-        raise NotImplementedError("The function does not work for type {}.".format(
-            self.knn_perm_.dtype))
+        raise NotImplementedError(  # pragma: no cover
+            "The function does not work for type {}.".format(
+                self.knn_perm_.dtype))
 
     def transform(self, X, y):
         """

mlinsights/mlmodel/target_predictors.py

@@ -109,7 +109,7 @@ def predict(self, X):
             Predicted values.
         """
         if not hasattr(self, 'regressor_'):
-            raise NotFittedError(
+            raise NotFittedError(  # pragma: no cover
                 "This instance {} is not fitted yet. Call 'fit' with "
                 "appropriate arguments before using this method.".format(
                     type(self)))

mlinsights/mltree/tree_structure.py

@@ -14,7 +14,8 @@ def _get_tree(obj):
         return obj
     if hasattr(obj, "tree_"):
         return obj.tree_
-    raise AttributeError("obj is no tree: {}".format(type(obj)))
+    raise AttributeError(  # pragma: no cover
+        "obj is no tree: {}".format(type(obj)))
 
 
 def tree_leave_index(model):