def test_launch_model_check_default_parameters(self, partial):
# Check default parameters.
param_config = {
"verbose": "no",
}
if partial:
param_config["model_parameters"] = {}
for n_threads in range(1, 7):
param_config["max_threads"] = n_threads
df = get_fake_df(10)
predictor = MockUpModel(param_config)
model_result = predictor.launch_model(df.copy(),
max_threads=n_threads)
expected_train_set_lengths = [2, 4, 6, 8, 9]
assert predictor.test_values == 1
assert predictor.delta_training_values == 2
assert predictor.main_accuracy_estimator == "mae"
assert type(predictor.transformation) == Identity
assert len(model_result.results) == 5
for r in model_result.results:
prediction = r.prediction
testing_performances = r.testing_performances
first_used_index = testing_performances.first_used_index
used_training_set = df.loc[first_used_index:]
used_training_set = used_training_set.iloc[:-1]
assert len(prediction) == len(used_training_set) + 1 + 10
expected_train_set_lengths.remove(len(used_training_set))
assert len(expected_train_set_lengths) == 0
def test_launch_model_with_extra_regressors(self):
# Test extra-regressors.
param_config = {
"model_parameters": {
"test_percentage": 10,
"delta_training_percentage": 20,
"prediction_lags": 10,
"transformation": "none",
"main_accuracy_estimator": "mae"
},
}
for n_threads in range(1, 7):
param_config["max_threads"] = n_threads
df = get_fake_df(10)
extra_regressor = get_fake_df(20, name="A")
predictor = MockUpModel(param_config)
model_result = predictor.launch_model(
df.copy(), extra_regressors=extra_regressor)
expected_train_set_lengths = [2, 4, 6, 8, 9]
assert predictor.test_values == 1
assert predictor.delta_training_values == 2
assert predictor.main_accuracy_estimator == "mae"
assert len(model_result.results) == 5
assert predictor.extra_regressors_in_predict is extra_regressor
assert predictor.extra_regressors_in_training is extra_regressor
for r in model_result.results:
prediction = r.prediction
testing_performances = r.testing_performances
first_used_index = testing_performances.first_used_index
used_training_set = df.loc[first_used_index:]
used_training_set = used_training_set.iloc[:-1]
assert len(prediction) == len(used_training_set) + 1 + 10
expected_train_set_lengths.remove(len(used_training_set))
assert len(expected_train_set_lengths) == 0
def test_launch_percentages_over_100(self):
# Percentages' sum is over 100%; adapt the window.
param_config = {
"model_parameters": {
"test_values": 5,
"delta_training_percentage": 100,
"prediction_lags": 10,
"transformation": "none",
"main_accuracy_estimator": "mae"
},
}
for n_threads in range(1, 3):
param_config["max_threads"] = n_threads
df = get_fake_df(100)
predictor = MockUpModel(param_config)
model_result = predictor.launch_model(df.copy(),
max_threads=n_threads)
expected_train_set_lengths = [95]
assert predictor.test_values == 5
assert predictor.delta_training_values == 100
assert predictor.main_accuracy_estimator == "mae"
assert len(model_result.results) == 1
for r in model_result.results:
prediction = r.prediction
testing_performances = r.testing_performances
first_used_index = testing_performances.first_used_index
used_training_set = df.loc[first_used_index:]
used_training_set = used_training_set.iloc[:-5]
assert len(prediction) == len(used_training_set) + 5 + 10
expected_train_set_lengths.remove(len(used_training_set))
assert len(expected_train_set_lengths) == 0
def test_launch_model_float_percentages(self):
# Percentages' sum is not 100% and values are float; adapt the windows. Use "test_percentage".
param_config = {
"model_parameters": {
"test_percentage": 11.7,
"delta_training_percentage": 18.9,
"prediction_lags": 10,
"transformation": "none",
"main_accuracy_estimator": "mae"
},
}
for n_threads in range(1, 7):
param_config["max_threads"] = n_threads
df = get_fake_df(101)
predictor = MockUpModel(param_config)
model_result = predictor.launch_model(df.copy(),
max_threads=n_threads)
expected_train_set_lengths = [19, 38, 57, 76, 89]
assert predictor.test_values == 12
assert predictor.delta_training_values == 19
assert predictor.main_accuracy_estimator == "mae"
assert len(model_result.results) == 5
for r in model_result.results:
prediction = r.prediction
testing_performances = r.testing_performances
first_used_index = testing_performances.first_used_index
used_training_set = df.loc[first_used_index:]
used_training_set = used_training_set.iloc[:-12]
assert len(prediction) == len(used_training_set) + 10 + 12
expected_train_set_lengths.remove(len(used_training_set))
assert len(expected_train_set_lengths) == 0