def test_backtester_test_metrics(iclaims_training_data, metrics):
df = iclaims_training_data
lgt = LGTMAP(response_col='claims',
date_col='week',
seasonality=1,
verbose=False)
backtester = BackTester(
model=lgt,
df=df,
forecast_len=3,
n_splits=1,
)
backtester.fit_predict()
eval_out = backtester.score(metrics=metrics)
evaluated_metrics = set(eval_out['metric_name'].tolist())
if metrics is None:
expected_metrics = [x.__name__ for x in backtester._default_metrics]
elif isinstance(metrics, list):
expected_metrics = [x.__name__ for x in metrics]
else:
expected_metrics = [metrics.__name__]
assert set(expected_metrics) == evaluated_metrics
def test_lgt_predict_all_positive_reg(iclaims_training_data):
df = iclaims_training_data
lgt = LGTMAP(
response_col='claims',
date_col='week',
regressor_col=['trend.unemploy', 'trend.filling', 'trend.job'],
regressor_sign=['+', '+', '+'],
seasonality=52,
seed=8888,
)
lgt.fit(df)
predicted_df = lgt.predict(df, decompose=True)
assert any(predicted_df['regression'].values)
def test_backtester_with_training_data(iclaims_training_data):
df = iclaims_training_data
lgt = LGTMAP(response_col='claims',
date_col='week',
seasonality=1,
verbose=False)
backtester = BackTester(
model=lgt,
df=df,
min_train_len=100,
incremental_len=100,
forecast_len=20,
)
backtester.fit_predict()
eval_out = backtester.score(include_training_metrics=True)
evaluated_test_metrics = set(eval_out.loc[~eval_out['is_training_metric'],
'metric_name'].tolist())
evaluated_train_metrics = set(eval_out.loc[eval_out['is_training_metric'],
'metric_name'].tolist())
expected_test_metrics = [x.__name__ for x in backtester._default_metrics]
expected_train_metrics = list(
filter(
lambda x: backtester._get_metric_callable_signature(x) ==
{'actual', 'predicted'}, backtester._default_metrics))
expected_train_metrics = [x.__name__ for x in expected_train_metrics]
assert set(expected_test_metrics) == evaluated_test_metrics
assert set(expected_train_metrics) == evaluated_train_metrics
def test_lgt_map_univariate(synthetic_data, estimator_type):
train_df, test_df, coef = synthetic_data
lgt = LGTMAP(response_col='response',
date_col='week',
seasonality=52,
verbose=False,
estimator_type=estimator_type)
lgt.fit(train_df)
predict_df = lgt.predict(test_df)
expected_columns = ['week', 'prediction']
expected_shape = (51, len(expected_columns))
expected_num_parameters = 12 # no `lp__` parameter in optimizing()
assert predict_df.shape == expected_shape
assert predict_df.columns.tolist() == expected_columns
assert len(lgt._posterior_samples) == expected_num_parameters
def test_lgt_map_single_regressor(iclaims_training_data):
df = iclaims_training_data
df['claims'] = np.log(df['claims'])
regressor_col = ['trend.unemploy']
lgt = LGTMAP(
response_col='claims',
date_col='week',
regressor_col=regressor_col,
seasonality=52,
seed=8888,
)
lgt.fit(df)
predicted_df = lgt.predict(df)
expected_num_parameters = 13
expected_columns = ['week', 'prediction_5', 'prediction', 'prediction_95']
assert predicted_df.shape[0] == df.shape[0]
assert predicted_df.columns.tolist() == expected_columns
assert len(lgt._posterior_samples) == expected_num_parameters
def test_lgt_fixed_sm_input(synthetic_data, level_sm_input,
seasonality_sm_input, slope_sm_input):
train_df, test_df, coef = synthetic_data
lgt = LGTMAP(
response_col='response',
date_col='week',
regressor_col=train_df.columns.tolist()[2:],
level_sm_input=level_sm_input,
seasonality_sm_input=seasonality_sm_input,
slope_sm_input=slope_sm_input,
seasonality=52,
verbose=False,
)
lgt.fit(train_df)
predict_df = lgt.predict(test_df, n_bootstrap_draw=100)
regression_out = lgt.get_regression_coefs()
num_regressors = regression_out.shape[0]
expected_columns = ['week', 'prediction_5', 'prediction', 'prediction_95']
expected_shape = (51, len(expected_columns))
expected_regression_shape = (6, 3)
assert predict_df.shape == expected_shape
assert predict_df.columns.tolist() == expected_columns
assert regression_out.shape == expected_regression_shape
assert num_regressors == len(train_df.columns.tolist()[2:])
def test_lgt_map_fit(synthetic_data, seasonality, estimator_type):
train_df, test_df, coef = synthetic_data
lgt = LGTMAP(response_col='response',
date_col='week',
seasonality=seasonality,
verbose=False,
estimator_type=estimator_type)
lgt.fit(train_df)
init_call = lgt.get_init_values()
if seasonality:
assert isinstance(init_call, LGTInitializer)
assert init_call.s == 52
init_values = init_call()
assert init_values['init_sea'].shape == (51, )
else:
assert not init_call
predict_df = lgt.predict(test_df)
expected_num_parameters = 10
expected_columns = ['week', 'prediction_5', 'prediction', 'prediction_95']
if seasonality == 52:
expected_num_parameters += 2
expected_shape = (51, len(expected_columns))
assert predict_df.shape == expected_shape
assert predict_df.columns.tolist() == expected_columns
assert len(lgt._posterior_samples) == expected_num_parameters
def test_lgt_map_reproducibility(synthetic_data, seasonality):
train_df, test_df, coef = synthetic_data
lgt1 = LGTMAP(
response_col='response',
date_col='week',
prediction_percentiles=[5, 95],
seasonality=seasonality,
)
# first fit and predict
lgt1.fit(train_df)
posteriors1 = copy(lgt1._aggregate_posteriors['map'])
prediction1 = lgt1.predict(test_df)
# second fit and predict
# note a new instance must be created to reset the seed
# note both fit and predict contain random generation processes
lgt2 = LGTMAP(
response_col='response',
date_col='week',
prediction_percentiles=[5, 95],
seasonality=seasonality,
)
lgt2.fit(train_df)
posteriors2 = copy(lgt2._aggregate_posteriors['map'])
prediction2 = lgt2.predict(test_df)
# assert same posterior keys
assert set(posteriors1.keys()) == set(posteriors2.keys())
# assert posterior draws are reproducible
for k, v in posteriors1.items():
assert np.allclose(posteriors1[k], posteriors2[k])
# assert prediction is reproducible
assert np.allclose(prediction1['prediction'].values,
prediction2['prediction'].values)
def test_lgt_map_fit(synthetic_data, seasonality, estimator_type):
train_df, test_df, coef = synthetic_data
lgt = LGTMAP(
response_col='response',
date_col='week',
seasonality=seasonality,
verbose=False,
estimator_type=estimator_type
)
lgt.fit(train_df)
predict_df = lgt.predict(test_df)
expected_num_parameters = 10
expected_columns = ['week', 'prediction_5', 'prediction', 'prediction_95']
if seasonality == 52:
expected_num_parameters += 2
expected_shape = (51, len(expected_columns))
assert predict_df.shape == expected_shape
assert predict_df.columns.tolist() == expected_columns
assert len(lgt._posterior_samples) == expected_num_parameters
def test_backtester_sceduler_args(iclaims_training_data, scheduler_args):
df = iclaims_training_data
lgt = LGTMAP(response_col='claims',
date_col='week',
seasonality=1,
verbose=False)
backtester = BackTester(
model=lgt,
df=df,
**scheduler_args,
)
backtester.fit_predict()
eval_out = backtester.score(metrics=[smape])
assert np.all(eval_out['metric_values'].values > 0)
def test_lgt_grid_tuning(synthetic_data, param_grid):
train_df, test_df, coef = synthetic_data
args = {'response_col': 'response', 'date_col': 'week', 'seasonality': 52}
lgt = LGTMAP(**args)
best_params, tuned_df = grid_search_orbit(param_grid,
model=lgt,
df=train_df,
min_train_len=80,
incremental_len=20,
forecast_len=20,
metrics=None,
criteria=None,
verbose=True)
assert best_params[0].keys() == param_grid.keys()
assert set(tuned_df.columns.to_list()) == set(
list(param_grid.keys()) + ['metrics'])
assert tuned_df.shape == (9, 3)
def test_backtester_test_data_only(iclaims_training_data):
df = iclaims_training_data
lgt = LGTMAP(response_col='claims',
date_col='week',
seasonality=1,
verbose=False)
backtester = BackTester(
model=lgt,
df=df,
min_train_len=100,
incremental_len=100,
forecast_len=20,
)
backtester.fit_predict()
eval_out = backtester.score()
evaluated_metrics = set(eval_out['metric_name'].tolist())
expected_metrics = [x.__name__ for x in backtester._default_metrics]
assert set(expected_metrics) == evaluated_metrics
def test_lgt_mixed_signs_and_order(iclaims_training_data, regressor_signs):
df = iclaims_training_data
df['claims'] = np.log(df['claims'])
raw_regressor_col = ['trend.unemploy', 'trend.filling', 'trend.job']
new_regressor_col = [raw_regressor_col[idx] for idx in [2, 1, 0]]
new_regressor_signs = [regressor_signs[idx] for idx in [2, 1, 0]]
# mixiing ordering of cols in df of prediction
new_df = df[['claims', 'week'] + new_regressor_col]
lgt = LGTMAP(
response_col='claims',
date_col='week',
regressor_col=raw_regressor_col,
regressor_sign=regressor_signs,
seasonality=52,
seed=8888,
)
lgt.fit(df)
predicted_df_v1 = lgt.predict(df)
predicted_df_v2 = lgt.predict(new_df)
# mixing ordering of signs
lgt_new = LGTMAP(
response_col='claims',
date_col='week',
regressor_col=new_regressor_col,
regressor_sign=new_regressor_signs,
seasonality=52,
seed=8888,
)
lgt_new.fit(df)
predicted_df_v3 = lgt_new.predict(df)
predicted_df_v4 = lgt_new.predict(new_df)
pred_v1 = predicted_df_v1['prediction'].values
pred_v2 = predicted_df_v2['prediction'].values
pred_v3 = predicted_df_v3['prediction'].values
pred_v4 = predicted_df_v4['prediction'].values
# they should be all identical; ordering of signs or columns in prediction show not matter
assert np.allclose(pred_v1, pred_v2, atol=1e-3)
assert np.allclose(pred_v1, pred_v3, atol=1e-3)
assert np.allclose(pred_v1, pred_v4, atol=1e-3)