def testConstructor_var_numpy(self):
m = orange.CostMatrix(3, self.num_array)
self.assertEqual(list(m), self.num_array.tolist())
m = orange.CostMatrix(self.class_var, self.num_array)
self.assertEqual(list(m), self.num_array.tolist())
self.assertRaises(IndexError, orange.CostMatrix, 4, self.num_array)
self.assertRaises(IndexError, orange.CostMatrix, 2, self.num_array)
self.assertRaises(IndexError, orange.CostMatrix, 3, numpy.zeros((3, 1)))
def testConstructor_dim_mat(self):
m = orange.CostMatrix(3, self.matrix)
self.assertEqual(list(m), self.matrix)
m = orange.CostMatrix(self.class_var, self.matrix)
self.assertEqual(list(m), self.matrix)
self.assertRaises(IndexError, orange.CostMatrix, 4, self.matrix)
self.assertRaises(IndexError, orange.CostMatrix, 2, self.matrix)
self.assertRaises(IndexError, orange.CostMatrix, 3, [[1], [2], [3]])
self.assertRaises(IndexError, orange.CostMatrix, 3, [[1, 2, 3, 4], [2], [3]])
def testPickle(self):
m = orange.CostMatrix(self.class_var, self.matrix)
s = pickle.dumps(m)
m2 = pickle.loads(s)
self.assertEqual(list(m2), self.matrix)
self.assertEqual(m2.dimension, m.dimension)
self.assertEqual(m2.class_var, self.class_var)
m = orange.CostMatrix(self.matrix)
s = pickle.dumps(m)
m2 = pickle.loads(s)
self.assertEqual(list(m2), self.matrix)
self.assertEqual(m2.dimension, m.dimension)
self.assertIsNone(m2.class_var)
def testIndexing(self):
m = orange.CostMatrix(self.class_var, self.matrix)
self.assertEqual(m[0, 0], self.matrix[0][0])
self.assertEqual(m[0, 2], self.matrix[0][2])
self.assertEqual(m[2, 0], self.matrix[2][0])
self.assertEqual(m[2, 2], self.matrix[2][2])
self.assertEqual(m[0], self.matrix[0])
self.assertEqual(m[2], self.matrix[2])
self.assertEqual(m[self.class_var.values[0]], self.matrix[0])
self.assertEqual(m[self.class_var.values[2]], self.matrix[2])
self.assertEqual(m[orange.PyValue(self.class_var, 0)], self.matrix[0])
self.assertEqual(m[orange.PyValue(self.class_var, 2)], self.matrix[2])
for i, j in ((0, 5), (5, 0), (0, -2), (-2, 0)):
with self.assertRaises(IndexError):
m[i, j]
with self.assertRaises(IndexError):
m[5]
with self.assertRaises(IndexError):
m[-2]
with self.assertRaises(IndexError):
m[1, 1, 1]
with self.assertRaises(ValueError):
m["foo"]
with self.assertRaises(TypeError):
m[orange]
with self.assertRaises(TypeError):
m[1:]
with self.assertRaises(TypeError):
m[1, 1:3]
def testConstructor_var(self):
m = orange.CostMatrix(self.class_var)
self.assertEqual(m.dimension, 3)
self.assertEqual(m.class_var, self.class_var)
for i in range(3):
for j in range(3):
self.assertEqual(m[i, j], i != j)
m = orange.CostMatrix(self.class_var, 42)
self.assertEqual(m.dimension, 3)
self.assertEqual(m.class_var, self.class_var)
for i in range(3):
for j in range(3):
self.assertEqual(m[i, j], 0 if i==j else 42)
self.assertRaises(ValueError, orange.CostMatrix, self.data.domain[0])
def testAssertIndex(self):
m = orange.CostMatrix(self.class_var, self.matrix)
m[1, 2] = 42
self.assertEqual(m[1, 2], 42)
m[1] = [3.14, 5.55, 1.11]
self.assertEqual(m[1], [3.14, 5.55, 1.11])
def testConstructor_dim(self):
m = orange.CostMatrix(3)
self.assertEqual(m.dimension, 3)
self.assertIsNone(m.class_var)
for i in range(3):
for j in range(3):
self.assertEqual(m[i, j], i != j)
m = orange.CostMatrix(3, 42)
self.assertEqual(m.dimension, 3)
self.assertIsNone(m.class_var)
for i in range(3):
for j in range(3):
self.assertEqual(m[i, j], 0 if i==j else 42)
self.assertRaises(ValueError, orange.CostMatrix, 0)
self.assertRaises(ValueError, orange.CostMatrix, -1)
self.assertRaises(TypeError, orange.CostMatrix, "wrogn!")
def testSetCost(self):
m = orange.CostMatrix(self.class_var, self.matrix)
m.setcost(0, 0, 42)
self.assertEqual(m.getcost(0, 0), 42)
m.setcost(0, 2, 3.14)
self.assertEqual(m.getcost(0, 2), 3.14)
m.setcost(self.class_var.values[0], self.class_var.values[2], 2.79)
self.assertEqual(m[self.class_var.values[0], self.class_var.values[2]], 2.79)
self.assertRaises(ValueError, m.setcost, "foo", "foo", 42)
self.assertRaises(TypeError, m.setcost, orange, 42, 42)
def testGetCost(self):
m = orange.CostMatrix(self.class_var, self.matrix)
self.assertEqual(m.getcost(0, 0), self.matrix[0][0])
self.assertEqual(m.getcost(0, 2), self.matrix[0][2])
self.assertEqual(m.getcost(2, 0), self.matrix[2][0])
self.assertEqual(
m.getcost(self.class_var.values[0], self.class_var.values[2]),
self.matrix[0][2])
for i, j in ((0, 5), (5, 0), (0, -2), (-2, 0)):
self.assertRaises(IndexError, m.getcost, i, j)
self.assertRaises(ValueError, m.getcost, "foo", "foo")
self.assertRaises(TypeError, m.getcost, orange, 42)
def testLen(self):
m = orange.CostMatrix(3)
self.assertEqual(len(m), 3)
def testIteration(self):
m = orange.CostMatrix(self.matrix)
for e, f in zip(m, self.matrix):
self.assertEqual(e, f)
# Description: Shows how to assess the quality of attributes not in the dataset
# Category: attribute quality
# Classes: EntropyDiscretization, MeasureAttribute, MeasureAttribute_info
# Uses: iris
# Referenced: MeasureAttribute.htm
import orange
print
print "Default matrix of size 3"
cm = orange.CostMatrix(3)
print "classVar =", cm.classVar
for pred in range(3):
for corr in range(3):
print cm.getcost(pred, corr),
print
print
print "Matrix for Iris, with default element 2 and several modified elements"
data = orange.ExampleTable("iris")
cm = orange.CostMatrix(data.domain.classVar, 2)
cm.setcost("Iris-setosa", "Iris-virginica", 1)
cm.setcost("Iris-versicolor", "Iris-virginica", 1)
print "classVar = %s, values = %s" % (cm.classVar.name, cm.classVar.values)
for pred in range(3):
for corr in range(3):
print cm.getcost(pred, corr),
print
print