def test_arima():
if os.path.exists("results"):
shutil.rmtree("results")
config = {
'directory': 'results/arima',
'freq': "30min",
'params': {
'arima_order': (1, 0, 0)
}
}
dataset = athena.Dataset("dfw_demand.csv.gz",
index="timestamp",
freq="30min",
max_days=500,
max_training_days=200,
predition_length=12,
test_start_values=["2019-07-27 13:00:00"],
test_sequence_length=4)
transform = DataTransformARIMA(['vehicles'])
evaluate_ARIMA(dataset, transform, config)
def get_dataset(config):
dataset = athena.Dataset(
config['filename'],
index=config['index'],
freq=config['freq'],
max_days=config['max_days'],
max_training_days=config['max_training_days'],
predition_length=config['prediction_length'],
test_start_values=config['test_start_values'],
test_sequence_length=config['test_sequence_length'])
return dataset
def evaluate_model(model, config):
if os.path.exists("results"):
shutil.rmtree("results")
dataset = athena.Dataset("dfw_demand.csv.gz",
index="timestamp",
freq="30min",
max_training_days=5,
predition_length=48,
test_start_values=["2019-07-27 00:00:00"],
test_sequence_length=1
)
transform = athena.utils.gluonts.DataTransformGluon(['vehicles'])
athena.utils.gluonts.evaluate_gluonts(dataset, transform, model, config)
df = pd.read_csv(glob.glob("{}/*.csv".format(config['directory']))[0])
assert len(df) == 48
def test_ds():
ds = athena.Dataset("dfw_demand.csv.gz",
index="timestamp",
freq="30min",
max_days=500,
max_training_days=200,
predition_length=1,
test_start_values=["2019-07-27 00:00:00"],
test_sequence_length=4
)
# Ensure the number of cv: len(test_start_values)*test_sequence_length
assert len(ds.cv) == 4
# Assert the max days in the dataset
assert len(ds.df) == 500*ds.rows_per_day
# verify each cv has a max of 200 training days
for test in ds.cv:
assert (test['train_stop'] - test['train_start'])/ds.rows_per_day <= 200
def evaluate_model(model, config):
if os.path.exists("results"):
shutil.rmtree("results")
dataset = athena.Dataset("dfw_demand.csv.gz",
index="timestamp",
freq="30min",
max_days=500,
max_training_days=200,
predition_length=1,
test_start_values=["2019-07-27 13:00:00"],
test_sequence_length=4)
transform = DataTransformSKLearn(
['vehicles'], ['vehicles_1', 'vehicles_2', 'vehicles_3'], ['hour'])
athena.utils.sklearn.evaluate_sklearn(dataset, transform, model, config)
df = pd.read_csv(glob.glob("{}/*.csv".format(config['directory']))[0])
assert len(df) == 4 #test_sequence_length*len(test_start_values)