Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

allow percentiles in bt output #724

Merged
merged 1 commit into from
Mar 9, 2022
Merged

allow percentiles in bt output #724

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
388 changes: 227 additions & 161 deletions examples/backtest.ipynb
View file
Open in desktop

Large diffs are not rendered by default.

9 changes: 5 additions & 4 deletions orbit/diagnostics/backtest.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -280,6 +280,7 @@ def fit_predict(self):
model_copy.fit(train_df)
train_predictions = model_copy.predict(train_df)
test_predictions = model_copy.predict(test_df)
all_pred_cols = [x for x in train_predictions.columns if x!= date_col]

# set attributes
self._fitted_models.append(model_copy)
Expand All @@ -297,13 +298,13 @@ def fit_predict(self):
train_dates = train_df[date_col].rename(BacktestFitKeys.DATE.value, axis='columns')
train_response = train_df[response_col].rename(BacktestFitKeys.ACTUAL.value, axis='columns')
train_values = pd.concat(
(train_dates, train_response, train_predictions[BacktestFitKeys.PREDICTED.value]), axis=1)
(train_dates, train_response, train_predictions[all_pred_cols]), axis=1)
train_values[BacktestFitKeys.TRAIN_FLAG.value] = True
# join test
test_dates = test_df[date_col].rename(BacktestFitKeys.DATE.value, axis='columns')
test_response = test_df[response_col].rename(BacktestFitKeys.ACTUAL.value, axis='columns')
test_values = pd.concat(
(test_dates, test_response, test_predictions[BacktestFitKeys.PREDICTED.value]), axis=1)
(test_dates, test_response, test_predictions[all_pred_cols]), axis=1)
test_values[BacktestFitKeys.TRAIN_FLAG.value] = False
# union train/test
both_values = pd.concat((train_values, test_values), axis=0)
Expand Down Expand Up @@ -341,9 +342,9 @@ def _validate_metric_callables(self, metrics):
if metric_signature == {BacktestFitKeys.ACTUAL.value, BacktestFitKeys.PREDICTED.value}:
continue
elif metric_signature.issubset({
BacktestFitKeys.TEST_ACTUAL.value,
BacktestFitKeys.TEST_ACTUAL.value,
BacktestFitKeys.TEST_PREDICTED.value,
BacktestFitKeys.TRAIN_ACTUAL.value,
BacktestFitKeys.TRAIN_ACTUAL.value,
BacktestFitKeys.TRAIN_PREDICTED.value
}):
continue
Expand Down
12 changes: 9 additions & 3 deletions tests/orbit/diagnostics/test_backtest.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@
import numpy as np

from orbit.diagnostics.backtest import TimeSeriesSplitter, BackTester
from orbit.diagnostics.metrics import smape, wmape, mape, mse, mae, rmsse
from orbit.diagnostics.metrics import smape, wmape, mae
from orbit.models import LGT, KTRLite

@pytest.mark.parametrize(
Expand Down Expand Up @@ -189,6 +189,8 @@ def test_backtester_with_training_data(iclaims_training_data):
date_col='week',
seasonality=1,
verbose=False,
n_bootstrap_draws=100,
prediction_percentiles=[10, 90],
estimator='stan-map'
)

Expand All @@ -201,12 +203,16 @@ def test_backtester_with_training_data(iclaims_training_data):
)

backtester.fit_predict()
out = backtester.get_predicted_df()
eval_out = backtester.score(include_training_metrics=True)

expected_out_columns = ['date', 'split_key', 'training_data', 'actual',
'prediction', 'prediction_10', 'prediction_90']
assert set(out.columns.tolist()) == set(expected_out_columns)

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', 'prediction'}, backtester._default_metrics)
)
Expand Down