plotting

plot_backtests(y_true, y_preds, *, n_series=10, seed=None, n_cols=2, last_n=DEFAULT_LAST_N, **kwargs)

Given panel DataFrame of observed values y and backtests across splits y_pred, returns subplots for each individual entity / time-series.

Parameters:

Name	Type	Description	Default
y_true	Union[DataFrame, LazyFrame]	Panel DataFrame of observed values.	required
y_preds	DataFrame	Panel DataFrame of backtested values.	required
n_series	int	Number of entities / time-series to plot. Defaults to 10.	10
seed	Union[int, None]	Random seed for sampling entities / time-series. Defaults to None.	None
n_cols	int	Number of columns to arrange subplots. Defaults to 2.	2
last_n	int	Plot last_n most recent values in y and y_pred. Defaults to 64.	DEFAULT_LAST_N

Returns:

Name	Type	Description
figure	Figure	Plotly subplots.

plot_comet(y_train, y_test, y_pred, scoring=None, **kwargs)

Given a train-test-split of panel data (y_train, y_test) and forecast y_pred, returns a Comet plot i.e. scatterplot of volatility per entity in y_train against the forecast scores.

Parameters:

Name	Type	Description	Default
y_train	DataFrame	Panel DataFrame of train dataset.	required
y_test	DataFrame	Panel DataFrame of test dataset.	required
y_pred	DataFrame	Panel DataFrame of forecasted values to score against y_test.	required
scoring	Optional[metric]	If None, defaults to SMAPE.	None

Returns:

Name	Type	Description
figure	Figure	Plotly scatterplot.

plot_entities(y, **kwargs)

Given panel DataFrame of observed values y, returns bar chart of entity counts, which is the number of observed values per entity.

Parameters:

Name	Type	Description	Default
y	DataFrame | LazyFrame	Panel DataFrame of observed values.	required

Returns:

Name	Type	Description
figure	Figure	Plotly bar chart.

plot_forecasts(y_true, y_pred, *, n_series=10, seed=None, n_cols=2, last_n=DEFAULT_LAST_N, **kwargs)

Given panel DataFrames of observed values y and forecasts y_pred, returns subplots for each individual entity / time-series.

Parameters:

Name	Type	Description	Default
y_true	Union[DataFrame, LazyFrame]	Panel DataFrame of observed values.	required
y_pred	DataFrame	Panel DataFrame of forecasted values.	required
n_series	int	Number of entities / time-series to plot. Defaults to 10.	10
seed	Union[int, None]	Random seed for sampling entities / time-series. Defaults to None.	None
n_cols	int	Number of columns to arrange subplots. Defaults to 2.	2
last_n	int	Plot last_n most recent values in y and y_pred. Defaults to 64.	DEFAULT_LAST_N

Returns:

Name	Type	Description
figure	Figure	Plotly subplots.

plot_fva(y_true, y_pred, y_pred_bench, scoring=None, **kwargs)

Given two panel data forecasts y_pred and y_pred_bench, returns scatterplot of benchmark scores against forecast scores. Each dot represents a single entity / time-series.

Parameters:

Name	Type	Description	Default
y_true	DataFrame	Panel DataFrame of test dataset.	required
y_pred	DataFrame	Panel DataFrame of forecasted values.	required
y_pred_bench	DataFrame	Panel DataFrame of benchmark forecast values.	required
scoring	Optional[metric]	If None, defaults to SMAPE.	None

Returns:

Name	Type	Description
figure	Figure	Plotly scatterplot.

plot_panel(y, *, n_series=10, seed=None, n_cols=2, last_n=DEFAULT_LAST_N, **kwargs)

Given panel DataFrames of observed values y, returns subplots for each individual entity / time-series.

Parameters:

Name	Type	Description	Default
y	Union[DataFrame, LazyFrame]	Panel DataFrame of observed values.	required
n_series	int	Number of entities / time-series to plot. Defaults to 10.	10
seed	Union[int, None]	Random seed for sampling entities / time-series. Defaults to None.	None
n_cols	int	Number of columns to arrange subplots. Defaults to 2.	2
last_n	int	Plot last_n most recent values in y and y_pred. Defaults to 64.	DEFAULT_LAST_N

Returns:

Name	Type	Description
figure	Figure	Plotly subplots.

plot_residuals(y_resids, n_bins=None, **kwargs)

Given panel DataFrame of residuals across splits y_resids, returns binned counts plot of forecast residuals colored by entity / time-series.

Useful for residuals analysis (bias and normality) at scale.

Parameters:

Name	Type	Description	Default
y_resids	Union[DataFrame, LazyFrame]	Panel DataFrame of forecast residuals (i.e. observed less forecast).	required
n_bins	int	Number of bins.	None

Returns:

Name	Type	Description
figure	Figure	Plotly histogram.