def forecasting_models_example():
"""
Build 3 models and evaluate the performance of model trees,
regression trees and standard linear regression.
"""
train_file = 'data/bikeSpeedVsIq_train.txt'
test_file = 'data/bikeSpeedVsIq_test.txt'
training_matrix = np.mat(utils.load_tsv_into_array(train_file))
test_matrix = np.mat(utils.load_tsv_into_array(test_file))
# training tree
tree = regression_trees.create_tree(training_matrix, ops=(1, 20))
y_hat = regression_trees.create_forecast(tree, test_matrix[:, 0])
accuracy = np.corrcoef(y_hat, test_matrix[:, 1], rowvar=0)[0, 1]
logging.info("training accuracy = {0}".format(accuracy))
# model tree
tree = regression_trees.create_tree(training_matrix,
regression_trees.model_leaf,
regression_trees.model_error, (1, 20))
y_hat = regression_trees.create_forecast(
tree, test_matrix[:, 0], regression_trees.model_tree_evaluation)
accuracy = np.corrcoef(y_hat, test_matrix[:, 1], rowvar=0)[0, 1]
logging.info("model tree accuracy = {0}".format(accuracy))
weights, x, y = regression_trees.linearly_solve(training_matrix)
for i in range(np.shape(test_matrix)[0]):
y_hat[i] = test_matrix[i, 0] * weights[1, 0] + weights[0, 0]
accuracy = np.corrcoef(y_hat, test_matrix[:, 1], rowvar=0)[0, 1]
logging.info("regression accuracy = {0}".format(accuracy))
def pruning_example():
complex_tree = more_complex_tree()
log_formatted_tree(complex_tree)
my_data = utils.load_tsv_into_array('data/ex2test.txt')
my_matrix = np.mat(my_data)
pruned = regression_trees.prune(complex_tree, my_matrix)
log_formatted_tree(pruned, 'pruned tree')
def pruning_example():
complex_tree = more_complex_tree()
log_formatted_tree(complex_tree)
my_data = utils.load_tsv_into_array('data/ex2test.txt')
my_matrix = np.mat(my_data)
pruned = regression_trees.prune(complex_tree, my_matrix)
log_formatted_tree(pruned, 'pruned tree')
def main():
data_array = utils.load_tsv_into_array('data/k_means_test_set.txt')
data_matrix = np.mat(data_array)
rand_cent = k_means.random_centroid(data_matrix, 2)
logging.info("random centroid = {rand_cent}".format(rand_cent=rand_cent))
euc = k_means.euclidean_distance(data_matrix[0], data_matrix[1])
logging.info("Euclidean distance = {euc}".format(euc=euc))
centroids, cluster_assignment = k_means.k_means(data_matrix, 4)
logging.info("centroids = {cent}".format(cent=centroids))
bisecting_k_means()
def test_k_means(self):
"""k_means - k means should build clusters"""
data_matrix = np.mat(utils.load_tsv_into_array('data/test_set_3.txt'))
centroids, cluster_assignment = k_means.k_means(data_matrix, 4)
expected = np.mat(
np.array([[1., 0.45675494], [0., 0.3032197], [3., 1.74481454],
[1., 0.80407696], [0., 1.02508049], [3., 2.59648559],
[1., 0.42859499], [0., 0.0305198], [3., 2.37924609],
[2., 0.], [0., 5.38984416], [3., 0.04519236],
[1., 1.23757291], [0., 0.01298907], [3., 3.28350116],
[1., 2.33205513], [0., 3.72839989], [3., 0.1398885],
[1., 0.03288099], [0., 0.4038706], [3., 1.00363352],
[1., 1.16346981], [0., 0.93928783], [3., 0.02261741],
[1., 3.42458409], [0., 5.92927609], [3., 0.98873759],
[1., 1.83018987], [0., 0.91125974], [3., 1.28677032]]))
self.assertEqual(cluster_assignment.any(), expected.any())
def test_k_means(self):
"""k_means - k means should build clusters"""
data_matrix = np.mat(utils.load_tsv_into_array('data/test_set_3.txt'))
centroids, cluster_assignment = k_means.k_means(data_matrix, 4)
expected = np.mat(np.array([[1., 0.45675494],
[0., 0.3032197],
[3., 1.74481454],
[1., 0.80407696],
[0., 1.02508049],
[3., 2.59648559],
[1., 0.42859499],
[0., 0.0305198],
[3., 2.37924609],
[2., 0.],
[0., 5.38984416],
[3., 0.04519236],
[1., 1.23757291],
[0., 0.01298907],
[3., 3.28350116],
[1., 2.33205513],
[0., 3.72839989],
[3., 0.1398885],
[1., 0.03288099],
[0., 0.4038706],
[3., 1.00363352],
[1., 1.16346981],
[0., 0.93928783],
[3., 0.02261741],
[1., 3.42458409],
[0., 5.92927609],
[3., 0.98873759],
[1., 1.83018987],
[0., 0.91125974],
[3., 1.28677032]]))
self.assertEqual(cluster_assignment.any(), expected.any())
def piecewise_linear_solve_example():
matrix_2 = np.mat(utils.load_tsv_into_array('data/exp2.txt'))
model_tree = regression_trees.create_tree(matrix_2,
regression_trees.model_leaf,
regression_trees.model_error)
log_formatted_tree(model_tree, 'model tree')
def more_complex_tree():
data = utils.load_tsv_into_array('data/ex0.txt')
matrix = np.mat(data)
tree = regression_trees.create_tree(matrix, ops=(0, 1))
return tree
def very_simple_tree():
data = utils.load_tsv_into_array('data/ex00.txt')
matrix = np.mat(data)
tree = regression_trees.create_tree(matrix)
log_formatted_tree(tree, "the tree")
def bisecting_k_means():
data_matrix = np.mat(utils.load_tsv_into_array('data/test_set_2.txt'))
centroid_list, assessments = k_means.bisect_k_means(data_matrix, 3)
return centroid_list, assessments
def piecewise_linear_solve_example():
matrix_2 = np.mat(utils.load_tsv_into_array('data/exp2.txt'))
model_tree = regression_trees.create_tree(matrix_2,
regression_trees.model_leaf,
regression_trees.model_error)
log_formatted_tree(model_tree, 'model tree')
def more_complex_tree():
data = utils.load_tsv_into_array('data/ex0.txt')
matrix = np.mat(data)
tree = regression_trees.create_tree(matrix, ops=(0, 1))
return tree
def very_simple_tree():
data = utils.load_tsv_into_array('data/ex00.txt')
matrix = np.mat(data)
tree = regression_trees.create_tree(matrix)
log_formatted_tree(tree, "the tree")