def test_random_forest_1(self):
test_dataset = Dataset(_dataset_name = 'watermelon_2.0', _dataset_file_path = './datasets/watermelon.csv')
test_dataset.load_dataset(verbose=False)
test_random_forest = RandomForest(10)
test_random_forest.set_dataset(test_dataset)
test_random_forest.generate_random_forest()
def test_select_best_attribute_to_split(self):
test_tree_node = TreeNode()
test_dataset = Dataset(_dataset_name = 'watermelon_2.0', _dataset_file_path = './datasets/watermelon.csv')
test_dataset.load_dataset(verbose=False)
test_tree_node.set_samples(test_dataset.samples)
test_tree_node.set_attribute_list(list(range(6)))
best_attribute_to_split = test_tree_node.select_best_attribute_to_split()
self.assertEqual(best_attribute_to_split,3)
def test_get_accuracy(self):
test_dataset = Dataset(_dataset_name = 'watermelon_3.0', _dataset_file_path = './datasets/watermelon2.csv')
test_dataset.load_dataset(verbose=False)
test_random_forest = RandomForest(10)
test_random_forest.set_dataset(test_dataset)
test_random_forest.generate_random_forest()
(test_random_forest.get_accuracy(test_dataset.samples))
def test_predict_batch(self):
test_dataset = Dataset(_dataset_name = 'watermelon_3.0', _dataset_file_path = './datasets/watermelon2.csv')
test_dataset.load_dataset(verbose=False)
test_random_forest = RandomForest(10)
test_random_forest.set_dataset(test_dataset)
test_random_forest.generate_random_forest()
predicted_result = test_random_forest.predict_batch(test_dataset.samples)
def test_split_by_attribute_internal(self):
test_tree_node = TreeNode()
test_dataset = Dataset(_dataset_name = 'watermelon_2.0', _dataset_file_path = './datasets/watermelon.csv')
test_dataset.load_dataset(verbose=False)
test_tree_node.set_samples(test_dataset.samples)
test_tree_node.set_attribute_list(list(range(6)))
attribute_values_samples_mapping_dict = test_tree_node.split_by_attribute_internal(3)
self.assertEqual(attribute_values_samples_mapping_dict[0].shape[0],9)
self.assertEqual(attribute_values_samples_mapping_dict[1].shape[0],5)
self.assertEqual(attribute_values_samples_mapping_dict[2].shape[0],3)
def test_generate_decision_tree_continuous(self):
test_dataset = Dataset(_dataset_name='watermelon_3.0',
_dataset_file_path='./datasets/watermelon2.csv')
test_dataset.load_dataset(verbose=False)
test_decision_tree = DecisionTree()
test_decision_tree.set_training_samples_root(test_dataset.samples)
test_decision_tree.set_attributes_list(
list(range(test_dataset.num_features)))
test_decision_tree.set_attributes_list(
test_dataset.feature_category_list)
decision_tree_root = test_decision_tree.generate_decision_tree(
test_decision_tree.training_samples_root,
test_decision_tree.attributes_list)
def test_generate_decision_tree(self):
test_dataset = Dataset(_dataset_name='watermelon_2.0',
_dataset_file_path='./datasets/watermelon.csv')
test_dataset.load_dataset(verbose=False)
test_decision_tree = DecisionTree()
test_decision_tree.set_training_samples_root(test_dataset.samples)
test_decision_tree.set_attributes_list(
test_dataset.feature_category_list)
decision_tree_root = test_decision_tree.generate_decision_tree(
test_decision_tree.training_samples_root,
test_decision_tree.attributes_list)
self.assertEqual(
len(decision_tree_root.child_node_list[0].child_node_list[1].
child_node_list), 3)
def test_vis_tree_1(self):
test_dataset = Dataset(_dataset_name='watermelon_2.0',
_dataset_file_path='./datasets/watermelon.csv')
test_dataset.load_dataset(verbose=False)
test_decision_tree = DecisionTree()
test_decision_tree.set_training_samples_root(test_dataset.samples)
test_decision_tree.set_attributes_list(
test_dataset.feature_category_list)
decision_tree_root = test_decision_tree.generate_decision_tree(
test_decision_tree.training_samples_root,
test_decision_tree.attributes_list)
test_decision_tree.set_root(decision_tree_root)
test_vis_tree = VisTree(test_decision_tree,test_dataset.feature2number_mapping,\
test_dataset.feature_name_list,_tree_name="test_decision_tree")
test_vis_tree.vis_tree(mode=1)
def test_generate_random_decision_tree_2(self):
test_dataset = Dataset(_dataset_name='watermelon_3.0',
_dataset_file_path='./datasets/watermelon2.csv')
test_dataset.load_dataset(verbose=False)
test_decision_tree = DecisionTree()
test_decision_tree.set_training_samples_root(test_dataset.samples)
test_decision_tree.set_attributes_list(
test_dataset.feature_category_list)
decision_tree_root_1 = test_decision_tree.generate_decision_tree(
test_decision_tree.training_samples_root,
test_decision_tree.attributes_list,
random_state=1)
decision_tree_root_2 = test_decision_tree.generate_decision_tree(
test_decision_tree.training_samples_root,
test_decision_tree.attributes_list,
random_state=2)
def test_get_all_possible_values_on_attribute(self):
test_dataset = Dataset(_dataset_name='watermelon_2.0',
_dataset_file_path='./datasets/watermelon.csv')
test_dataset.load_dataset(verbose=False)
test_decision_tree = DecisionTree()
test_decision_tree.set_training_samples_root(test_dataset.samples)
test_decision_tree.set_attributes_list(
test_dataset.feature_category_list)
self.assertEqual(
test_decision_tree.get_all_possible_values_on_attribute((0, 0)),
[0, 1, 2])
self.assertEqual(
test_decision_tree.get_all_possible_values_on_attribute((1, 0)),
[0, 1, 2])
self.assertEqual(
test_decision_tree.get_all_possible_values_on_attribute((5, 0)),
[0, 1])
def test_get_ent(self):
test_tree_node = TreeNode()
ent1 = test_tree_node.get_ent(np.ones((4,5)))
self.assertEqual(ent1,0)
ent2 = test_tree_node.get_ent(np.zeros((4,5)))
self.assertEqual(ent2,0)
input_np = np.ones((4,5))
input_np[0][4] = 0
input_np[1][4] = 0
ent2 = test_tree_node.get_ent(input_np)
self.assertEqual(ent2,1)
test_dataset = Dataset(_dataset_name = 'watermelon_2.0', _dataset_file_path = './datasets/watermelon.csv')
test_dataset.load_dataset(verbose=False)
ent3 = test_tree_node.get_ent(test_dataset.samples)
self.assertAlmostEqual(ent3,0.9975025463691152)
def test_random_forest_3(self):
test_dataset = Dataset(_dataset_name = 'uci_blood', _dataset_file_path = './datasets/uci_blood.csv')
test_dataset.load_dataset(verbose=False)
test_random_forest = RandomForest(n_estimators = 20,n_samples=400)
test_random_forest.set_dataset(test_dataset)
import time
start = time.clock()
test_random_forest.generate_random_forest()
end = time.clock()
print((end-start)/20.0)
""" i = 0
for tree in test_random_forest.forest:
test_vis_tree = VisTree(tree,test_dataset.feature2number_mapping,\
test_dataset.feature_name_list,_tree_name="test_3_random_decision_tree_%d" %(i))
test_vis_tree.vis_tree(mode=1)
i += 1
"""
print(test_random_forest.calculate_out_of_bag_error())
def test_decision_tree_predict(self):
test_dataset = Dataset(_dataset_name='watermelon_3.0',
_dataset_file_path='./datasets/watermelon2.csv')
test_dataset.load_dataset(verbose=False)
test_decision_tree = DecisionTree()
test_decision_tree.set_training_samples_root(test_dataset.samples)
test_decision_tree.set_attributes_list(
test_dataset.feature_category_list)
decision_tree_root = test_decision_tree.generate_decision_tree(
test_decision_tree.training_samples_root,
test_decision_tree.attributes_list)
test_decision_tree.set_root(decision_tree_root)
for i in range((test_dataset.num_samples)):
test_sample = test_dataset.samples[i, :]
test_X = test_sample[0:-1]
test_y = test_sample[-1]
test_predicted_label = test_decision_tree.predict(
test_sample=test_X)
self.assertEqual(test_predicted_label, test_y)
def test_random_forest_predict(self):
test_dataset = Dataset(_dataset_name = 'watermelon_3.0', _dataset_file_path = './datasets/watermelon2.csv')
test_dataset.load_dataset(verbose=False)
test_random_forest = RandomForest(10)
test_random_forest.set_dataset(test_dataset)
test_random_forest.generate_random_forest()
# i = 0
# for tree in test_random_forest.forest:
# test_vis_tree = VisTree(tree,test_dataset.feature2number_mapping,\
# test_dataset.feature_name_list,_tree_name="test_2_random_decision_tree_%d" %(i))
# test_vis_tree.vis_tree(mode=1)
# i += 1
for i in range((test_dataset.num_samples)):
test_sample = test_dataset.samples[i,:]
test_X = test_sample[0:-1]
test_y = test_sample[-1]
test_predicted_label = test_random_forest.predict(test_sample=test_X)
import graphviz
# dot_data = tree.export_graphviz(clf, out_file="./random_forests_sklearn/vis/iris")
# graph = graphviz.Source(dot_data)
# graphviz.render(engine="dot",format="pdf",filepath="./random_forests_sklearn/vis/iris")
from random_forests.utils import Dataset
# test_dataset = Dataset(_dataset_name = 'watermelon_2.0', _dataset_file_path = './datasets/watermelon.csv')
# test_dataset.load_dataset(verbose=False)
# X = test_dataset.samples[:,:-1].astype(int)
# y = test_dataset.labels.astype(int)
# clf = tree.DecisionTreeClassifier(criterion="entropy")
# clf = clf.fit(X,y)
test_dataset = Dataset(_dataset_name='uci_blood',
_dataset_file_path='./datasets/uci_blood.csv')
test_dataset.load_dataset(verbose=False)
X = test_dataset.samples[:, :-1].astype(int)
y = test_dataset.labels.astype(int)
clf = tree.DecisionTreeClassifier(criterion="entropy")
stupid_clf = DummyClassifier(strategy='uniform')
RF_clf = RandomForestClassifier(n_estimators=200)
clf = clf.fit(X, y)
RF_clf = RF_clf.fit(X, y)
aver_list = cross_val_score(clf, X, y, cv=5, scoring='accuracy')
aver_list2 = cross_val_score(stupid_clf, X, y, cv=5, scoring='accuracy')
aver_list3 = cross_val_score(RF_clf, X, y, cv=5, scoring='accuracy')
print(aver_list)
print(aver_list2)
def main():
test_dataset = Dataset(_dataset_name='watermelon_2.0',
_dataset_file_path='./datasets/watermelon.csv')
test_dataset.load_dataset()
myDecisionTree = DecisionTree()