Merge pull request #5411 from PrimozGodec/fix-distributions

Distributions: fix bins that values on the edge fall in the right bin

Author: janezd

Orange/preprocess/discretize.py

@@ -320,8 +320,11 @@ def decimal_binnings(
         nbins = np.round((mx_ - mn_) / width)
         if min_bins <= nbins <= max_bins \
                 and (not bins or bins[-1].nbins != nbins):
-            bin_def = BinDefinition(mn_ + width * np.arange(nbins + 1),
-                                    label_fmt, None, width)
+            bins_ = mn_ + width * np.arange(nbins + 1)
+            # to prevent values on the edge of the bin fall in the wrong bin
+            # due to precision error on decimals that are not precise
+            bins_ = np.around(bins_, decimals=np.finfo(bins_.dtype).precision)
+            bin_def = BinDefinition(bins_, label_fmt, None, width)
             bins.append(bin_def)
     return bins
 

Orange/tests/test_discretize.py

@@ -8,7 +8,7 @@
 import numpy as np
 import scipy.sparse as sp
 
-from Orange.preprocess import discretize, Discretize
+from Orange.preprocess import discretize, Discretize, decimal_binnings
 from Orange import data
 from Orange.data import Table, Instance, Domain, ContinuousVariable, DiscreteVariable
 
@@ -95,6 +95,34 @@ def test_equalwidth_const_value(self):
         self.assertEqual(dvar.compute_value.points, [])
 
 
+class TestBinning(TestCase):
+    def test_decimal_binnings(self):
+        values = np.array([
+            -0.2, -0.2, -0.6, 1.0, 0.2, -0.6, 0.6, 1.0, 0.4, -0.5, -0.4, -0.4,
+            -0.6, 0.6, 0.75, 0.4, -0.2, 0.2, 0.0, 0.0, -1.0, -0.6, -0.2, -0.6,
+        ])
+        binning = decimal_binnings(values, factors=[0.2, 0.25, 0.5])
+        self.assertEqual(len(binning), 3)
+
+        np.testing.assert_array_equal(
+            binning[0].thresholds,
+            [-1, -0.8, -0.6, -0.4, -0.2, 0, 0.2, 0.4, 0.6, 0.8, 1]
+        )
+        self.assertEqual(binning[0].width, 0.2)
+
+        np.testing.assert_array_equal(
+            binning[1].thresholds,
+            [-1, -0.75, -0.5, -0.25, 0, 0.25, 0.5, 0.75, 1]
+        )
+        self.assertEqual(binning[1].width, 0.25)
+
+        np.testing.assert_array_equal(
+            binning[2].thresholds,
+            [-1, -0.5, 0, 0.5, 1]
+        )
+        self.assertEqual(binning[2].width, 0.5)
+
+
 # noinspection PyPep8Naming
 class TestEntropyMDL(TestCase):
     def test_entropy_with_two_values(self):