Visualization utilities for the benchmarking pipeline.

Visualizer

chronax.visualizer.Visualizer

Creates visualizations for time series forecasting results. The Visualizer provides methods for prediction plots (Matplotlib) and logging forecast comparisons to TensorBoard as scalar series (no PNGs).

__init__(self)

Initializes the Visualizer, setting Matplotlib and Seaborn styles.

Parameter	Type	Default	Description
self	-	-	(undocumented)

plot_predictions(self, y_true, y_pred, time_index=None, y_pred_lower=None, y_pred_upper=None, title='Predictions vs Actual Values', save_path=None)

(undocumented)

Parameter	Type	Default	Description
y_true	Union[pd.Series, np.ndarray]	-	(undocumented)
y_pred	Union[pd.Series, np.ndarray]	-	(undocumented)
time_index	Optional[Union[pd.DatetimeIndex, list]]	None	(undocumented)
y_pred_lower	Optional[Union[pd.Series, np.ndarray]]	None	(undocumented)
y_pred_upper	Optional[Union[pd.Series, np.ndarray]]	None	(undocumented)
title	str	"Predictions vs Actual Values"	(undocumented)
save_path	Optional[str]	None	(undocumented)

plot_residuals(self, y_true, y_pred, title='Residual Analysis', save_path=None)

(undocumented)

Parameter	Type	Default	Description
y_true	Union[pd.Series, np.ndarray]	-	(undocumented)
y_pred	Union[pd.Series, np.ndarray]	-	(undocumented)
title	str	"Residual Analysis"	(undocumented)
save_path	Optional[str]	None	(undocumented)

plot_forecast_window(self, *, y_pred: np.ndarray, y_true: np.ndarray, timestamps_pred: np.ndarray, model_name: str, task_name: str, hyperparameters: Optional[Mapping[str, object]] = None, window_idx: int = 0) -> None

Log actual vs predicted as TensorBoard scalars (no image files).

See :meth:LogManager.log_forecast_window_scalars. Tags group by model, then task, forecast origin (first timestamps_pred), and a stable hyperparameter trial segment when tuning so multiple grid points do not collide in TensorBoard.

Parameter	Type	Default	Description
y_pred	np.ndarray	-	(undocumented)
y_true	np.ndarray	-	(undocumented)
timestamps_pred	np.ndarray	-	(undocumented)
model_name	str	-	(undocumented)
task_name	str	-	(undocumented)
hyperparameters	Optional[Mapping[str, object]]	None	Params used to produce y_true/y_pred (omit or {} for a single-configuration model).
window_idx	int	0	(undocumented)

Raises: * ValueError: If y_true, y_pred, or timestamps_pred have incorrect shapes.