Fixes #73, fix __repr__ for ClassifierKMeans

sdpython · sdpython · commit 7ee31d0e462e · 2020-05-15T00:01:05.000+02:00
diff --git a/_unittests/ut_mlmodel/test_classification_kmeans.py b/_unittests/ut_mlmodel/test_classification_kmeans.py
@@ -1,14 +1,16 @@
 # -*- coding: utf-8 -*-
 """
-@brief      test log(time=2s)
+@brief      test log(time=20s)
 """
 import unittest
 import numpy
 from numpy.random import RandomState
 from sklearn import datasets
 from pyquickhelper.pycode import ExtTestCase
-from mlinsights.mlmodel import ClassifierAfterKMeans
-from mlinsights.mlmodel import test_sklearn_pickle, test_sklearn_clone, test_sklearn_grid_search_cv
+from mlinsights.mlmodel import (
+    ClassifierAfterKMeans, test_sklearn_pickle, test_sklearn_clone,
+    test_sklearn_grid_search_cv
+)
 
 
 class TestClassifierAfterKMeans(ExtTestCase):
@@ -73,6 +75,16 @@ def test_classification_kmeans_relevance(self):
         score = clk.score(X, Y)
         self.assertGreater(score, 0.95)
 
+    def test_issue(self):
+        X, labels_true = datasets.make_blobs(
+            n_samples=750, centers=6, cluster_std=0.4)
+        labels_true = labels_true % 3
+        clcl = ClassifierAfterKMeans(e_max_iter=1000)
+        clcl.fit(X, labels_true)
+        r = repr(clcl)
+        self.assertIn('ClassifierAfterKMeans(', r)
+        self.assertIn("c_init='k-means++'", r)
+
 
 if __name__ == "__main__":
     unittest.main()
diff --git a/_unittests/ut_mlmodel/test_kmeans_l1.py b/_unittests/ut_mlmodel/test_kmeans_l1.py
@@ -1,6 +1,6 @@
 # -*- coding: utf-8 -*-
 """
-@brief      test log(time=2s)
+@brief      test log(time=6s)
 """
 import unittest
 import numpy
diff --git a/_unittests/ut_mlmodel/test_kmeans_sklearn.py b/_unittests/ut_mlmodel/test_kmeans_sklearn.py
@@ -1,6 +1,6 @@
 # -*- coding: utf-8 -*-
 """
-@brief      test log(time=2s)
+@brief      test log(time=5s)
 """
 import unittest
 import numpy as np
diff --git a/mlinsights/__init__.py b/mlinsights/__init__.py
@@ -4,7 +4,7 @@
 @brief Module *mlinsights*.
 Look for insights for machine learned models.
 """
-__version__ = "0.2.420"
+__version__ = "0.2.422"
 __author__ = "Xavier Dupré"
 __github__ = "https://github.com/sdpython/mlinsights"
 __url__ = "http://www.xavierdupre.fr/app/mlinsights/helpsphinx/index.html"
diff --git a/mlinsights/mlmodel/classification_kmeans.py b/mlinsights/mlmodel/classification_kmeans.py
@@ -2,6 +2,7 @@
 @file
 @brief Combines a *k-means* followed by a predictor.
 """
+import textwrap
 import inspect
 import numpy
 from sklearn.linear_model import LogisticRegression
@@ -134,7 +135,8 @@ def get_params(self, deep=True):
         @param      deep        unused here
         @return                 dict
 
-        :meth:`set_params <mlinsights.mlmodel.classification_kmeans.ClassifierAfterKMeans.set_params>`
+        :meth:`set_params <mlinsights.mlmodel.classification_kmeans.
+        ClassifierAfterKMeans.set_params>`
         describes the pattern parameters names follow.
         """
         res = {}
@@ -164,3 +166,14 @@ def set_params(self, **values):
                 raise ValueError("Unexpected parameter name '{0}'".format(k))
         self.clus.set_params(**pc)
         self.estimator.set_params(**pe)
+
+    def __repr__(self):
+        """
+        Overloads `repr` as *scikit-learn* now relies
+        on the constructor signature.
+        """
+        el = ', '.join(['%s=%r' % (k, v)
+                        for k, v in self.get_params().items()])
+        text = "%s(%s)" % (self.__class__.__name__, el)
+        lines = textwrap.wrap(text, subsequent_indent='    ')
+        return "\n".join(lines)
