def test_linear_stacking_evaluation():
max_clicks = 3
step = 100
models = ["svr", "lars", "elastic", "perceptron"]
res_folder = "res"
metrics_folder = "metrics"
linear_predictions_folders = [
"{}_linear_predictions".format(model) for model in models
]
make_days_data()
os.mkdir(res_folder)
os.mkdir(metrics_folder)
for folder in linear_predictions_folders:
os.mkdir(folder)
for model, predictions_folder in zip(models, linear_predictions_folders):
train_linear(
os.path.join(new_pool_code_path, "calculating_predictions/run.py"),
days_data_path, predictions_folder, model, step)
train_linear_stacking(
os.path.join(new_pool_code_path, "calculating_predictions/run.py"),
days_data_path, res_folder, linear_predictions_folders, max_clicks)
os.system(
"python2 {} --data_folder {} --predictions_folder {} --out_folder {} --type argmax_classification"
.format(
os.path.join(new_pool_code_path, "evaluation/run_evaluation.py"),
days_data_path, res_folder, metrics_folder))
assert len([
line for line in open(os.path.join(metrics_folder, "metrics.txt"))
]) == DAYS_NUMBER - 2
clear([res_folder, days_data_path, metrics_folder] +
linear_predictions_folders)
def test_linear_stacking():
max_clicks = 3
step = 100
models = ["svr", "lars", "elastic", "perceptron"]
res_folder = "res"
linear_predictions_folders = [
"{}_linear_predictions".format(model) for model in models
]
make_days_data()
os.mkdir(res_folder)
for folder in linear_predictions_folders:
os.mkdir(folder)
for model, predictions_folder in zip(models, linear_predictions_folders):
train_linear(
os.path.join(new_pool_code_path, "calculating_predictions/run.py"),
days_data_path, predictions_folder, model, step)
train_linear_stacking(
os.path.join(new_pool_code_path, "calculating_predictions/run.py"),
days_data_path, res_folder, linear_predictions_folders, max_clicks)
assert len(os.listdir(res_folder)) == DAYS_NUMBER - 2
for i, filename in enumerate(os.listdir(res_folder)):
assert "train_{}_test_{}".format(i + 1, i + 2)
predictions = np.load(os.path.join(res_folder, filename))
assert len(np.shape(predictions)) == 3
assert np.shape(predictions)[1:] == (len(POSITIONS_VARIANTS),
max_clicks + 2)
clear([res_folder, days_data_path] + linear_predictions_folders)
def test_days_data():
make_days_data()
assert len(os.listdir(days_data_path)) == DAYS_NUMBER
for filename in os.listdir(days_data_path):
filename = os.path.join(days_data_path, filename)
with open(filename) as handler:
assert len([line for line in handler]) == 2
clear([days_data_path])
def test_regression():
make_days_data()
res_folder = "res"
os.mkdir(res_folder)
train_regression(
os.path.join(new_pool_code_path, "calculating_predictions/run.py"),
days_data_path, res_folder)
assert len(os.listdir(res_folder)) == DAYS_NUMBER - 1
for i, filename in enumerate(os.listdir(res_folder)):
assert "train_{}_test_{}".format(i, i + 1)
predictions = np.load(os.path.join(res_folder, filename))
assert len(np.shape(predictions)) == 2
assert np.shape(predictions)[1] == len(POSITIONS_VARIANTS)
clear([res_folder, days_data_path])
def test_binary_classification():
make_days_data()
res_folder = "res"
threshold = 0
os.mkdir(res_folder)
train_binary_classification(
os.path.join(new_pool_code_path, "calculating_predictions/run.py"),
days_data_path, res_folder, threshold)
assert len(os.listdir(res_folder)) == DAYS_NUMBER - 1
for i, filename in enumerate(os.listdir(res_folder)):
assert "train_{}_test_{}".format(i, i + 1)
predictions = np.load(os.path.join(res_folder, filename))
assert len(np.shape(predictions)) == 2
assert np.shape(predictions)[1] == 2
clear([res_folder, days_data_path])
def test_regression_evaluation():
make_days_data()
res_folder = "res"
metrics_folder = "metrics"
os.mkdir(res_folder)
os.mkdir(metrics_folder)
train_regression(
os.path.join(new_pool_code_path, "calculating_predictions/run.py"),
days_data_path, res_folder)
os.system(
"python2 {} --data_folder {} --predictions_folder {} --out_folder {} --type argmax_regression"
.format(
os.path.join(new_pool_code_path, "evaluation/run_evaluation.py"),
days_data_path, res_folder, metrics_folder))
assert len([
line for line in open(os.path.join(metrics_folder, "metrics.txt"))
]) == DAYS_NUMBER - 1
clear([res_folder, days_data_path, metrics_folder])
def test_linear():
step = 100
models = ["svr", "lars", "elastic", "perceptron"]
res_folder = "res"
linear_predictions_folders = [
"{}_linear_predictions".format(model) for model in models
]
make_days_data()
for folder in linear_predictions_folders:
os.mkdir(folder)
for model, res_folder in zip(models, linear_predictions_folders):
train_linear(
os.path.join(new_pool_code_path, "calculating_predictions/run.py"),
days_data_path, res_folder, model, step)
assert len(os.listdir(res_folder)) > 0
for folder in os.listdir(res_folder):
directory = os.path.join(res_folder, folder)
assert len(os.listdir(directory)) == DAYS_NUMBER - 1
for i, filename in enumerate(os.listdir(directory)):
assert "train_{}_test_{}".format(i, i + 1)
predictions = np.load(os.path.join(directory, filename))
assert len(np.shape(predictions)) == 1
clear([days_data_path] + linear_predictions_folders)
#!/bin/python2
from test_utils import clear, days_data_path
if __name__ == "__main__":
clear([days_data_path, "res", "metrics", "linear_predictions"])