def test_basic_tree(self):
tree = tree_lib.Tree(
node_lib.NonLeafNode(
condition=condition_lib.NumericalHigherThanCondition(
feature=dataspec_lib.SimpleColumnSpec(
name="f1", type=dataspec_lib.ColumnType.NUMERICAL),
threshold=1.5,
missing_evaluation=False),
pos_child=node_lib.LeafNode(value=value_lib.RegressionValue(
value=5.0, num_examples=10, standard_deviation=1.0)),
neg_child=node_lib.LeafNode(value=value_lib.ProbabilityValue(
probability=[0.5, 0.4, 0.1], num_examples=10))))
tree_repr = repr(tree)
logging.info("Tree repr:\n%s", tree_repr)
# The "repr" is a single line that does not contain any line return.
self.assertNotIn("\n", tree_repr)
logging.info("Tree str:\n%s", tree)
pretty = tree.pretty()
logging.info("Pretty:\n%s", pretty)
self.assertEqual(
pretty, """(f1 >= 1.5; miss=False, score=None)
├─(pos)─ RegressionValue(value=5.0,sd=1.0,n=10)
└─(neg)─ ProbabilityValue([0.5, 0.4, 0.1],n=10)
""")
def test_core_value_to_value_classifier(self):
core_node = decision_tree_pb2.Node()
core_node.classifier.distribution.counts[:] = [0.0, 8.0, 2.0]
core_node.classifier.distribution.sum = 10.0
self.assertEqual(
value_lib.core_value_to_value(core_node),
value_lib.ProbabilityValue(probability=[0.8, 0.2],
num_examples=10))
def test_non_leaf_with_children(self):
node = node_lib.NonLeafNode(
condition=condition_lib.NumericalHigherThanCondition(
feature=dataspec_lib.SimpleColumnSpec(
name="f1", type=dataspec_lib.ColumnType.NUMERICAL),
threshold=1.5,
missing_evaluation=False),
pos_child=node_lib.LeafNode(value=value_lib.RegressionValue(
value=5.0, num_examples=10, standard_deviation=1.0)),
neg_child=node_lib.LeafNode(value=value_lib.ProbabilityValue(
probability=[0.5, 0.4, 0.1], num_examples=10)))
logging.info("node:\n%s", node)
def test_basic_tree_with_label_classes(self):
tree = tree_lib.Tree(node_lib.NonLeafNode(
condition=condition_lib.NumericalHigherThanCondition(
feature=dataspec_lib.SimpleColumnSpec(
name="f1", type=dataspec_lib.ColumnType.NUMERICAL),
threshold=1.5,
missing_evaluation=False),
pos_child=node_lib.LeafNode(value=value_lib.RegressionValue(
value=5.0, num_examples=10, standard_deviation=1.0)),
neg_child=node_lib.LeafNode(value=value_lib.ProbabilityValue(
probability=[0.5, 0.4, 0.1], num_examples=10))),
label_classes=["x", "y", "z"])
plot = model_plotter.plot_tree(tree=tree)
self._save_plot(plot)
def test_probability(self):
value = value_lib.ProbabilityValue(probability=[0.5, 0.4, 0.1],
num_examples=10)
logging.info("value:\n%s", value)
def test_stump_with_label_classes(self):
tree = tree_lib.Tree(
node_lib.LeafNode(value=value_lib.ProbabilityValue(
probability=[0.5, 0.4, 0.1], num_examples=10)),
label_classes=["a", "b", "c"])
logging.info("Tree:\n%s", tree)