validate_forecast_sanity
chronax.utils.validate_forecast_sanity
Run sanity checks on forecasts to detect obvious issues.
validate_forecast_sanity(y_true, y_pred, point_forecast_statistic='mean') -> dict
| Parameter | Type | Default | Description |
|---|---|---|---|
y_true |
np.ndarray |
- | (undocumented) |
y_pred |
np.ndarray |
- | (undocumented) |
point_forecast_statistic |
str |
"mean" |
(undocumented) |
Returns: dict (a dict with keys: ok, warnings, stats.)
compute_point_forecast
chronax.utils.compute_point_forecast
Compute the point forecast from the predicted samples.
compute_point_forecast(y_pred, point_forecast_statistic) -> np.ndarray
| Parameter | Type | Default | Description |
|---|---|---|---|
y_pred |
np.ndarray |
- | Prediction array with shape (num_samples, time_steps, num_targets) for stochastic or (time_steps, num_targets) for deterministic. |
point_forecast_statistic |
str |
- | The statistic to apply. Supported values: "mean" (Compute mean across the first dimension (samples)) or "median" (Compute median across the first dimension (samples)). |
Returns: np.ndarray (Point forecast array with shape (time_steps, num_targets))
Raises:
* ValueError: If point_forecast_statistic is not supported