plots

Functions

metric_per_segment_distribution_plot(...[, ...])

Plot per-segment metrics distribution.

plot_backtest(forecast_df, ts[, segments, ...])

Plot targets and forecast for backtest pipeline.

plot_backtest_interactive(forecast_df, ts[, ...])

Plot targets and forecast for backtest pipeline using plotly.

plot_forecast(forecast_ts[, test_ts, ...])

Plot of prediction for forecast pipeline.

plot_forecast_decomposition(forecast_ts[, ...])

Plot of prediction and its components.

plot_metric_per_segment(metrics_df, metric_name)

Plot barplot with per-segment metrics.

plot_residuals(forecast_df, ts[, feature, ...])

Plot residuals for predictions from backtest against some feature.

prediction_actual_scatter_plot(forecast_df, ts)

Plot scatter plot with forecasted/actual values for segments.

qq_plot(residuals_ts[, qq_plot_params, ...])

Plot Q-Q plots for segments.

Classes

ComponentsMode(value)

Enum for components plotting modes.

MetricPlotType(value)

Enum for types of plot in metric_per_segment_distribution_plot().

PerFoldAggregation(value)

Enum for types of aggregation in a metric per-segment plot.
class ComponentsMode(value)[source]

Enum for components plotting modes.

class MetricPlotType(value)[source]

Enum for types of plot in metric_per_segment_distribution_plot().

hist

Histogram plot, seaborn.histplot() is used

box

Boxplot, seaborn.boxplot() is used

violin

Violin plot, seaborn.violinplot() is used

get_function()[source]

Get aggregation function.

class PerFoldAggregation(value)[source]

Enum for types of aggregation in a metric per-segment plot.

get_function()[source]

Get aggregation function.

metric_per_segment_distribution_plot(metrics_df: pandas.core.frame.DataFrame, metric_name: str, per_fold_aggregation_mode: Optional[str] = None, plot_type: Union[Literal['hist'], Literal['box'], Literal['violin']] = 'hist', seaborn_params: Optional[Dict[str, Any]] = None, figsize: Tuple[int, int] = (10, 5))[source]

Plot per-segment metrics distribution.

Parameters

  • metrics_df (pandas.core.frame.DataFrame) – dataframe with metrics calculated on the backtest

  • metric_name (str) – name of the metric to visualize

  • per_fold_aggregation_mode (Optional[str]) –

    • If None, separate distributions for each fold will be drawn

    • If str, determines how to aggregate metrics over the folds if they aren’t already aggregated

    (see PerFoldAggregation)

  • plot_type (Union[Literal['hist'], typing.Literal['box'], typing.Literal['violin']]) – type of plot (see MetricPlotType)

  • seaborn_params (Optional[Dict[str, Any]]) – dictionary with parameters for plotting

  • figsize (Tuple[int, int]) – size of the figure per subplot with one segment in inches

Raises

  • ValueError: – if metric_name isn’t present in metrics_df

  • NotImplementedError: – unknown per_fold_aggregation_mode is given

plot_backtest(forecast_df: pandas.core.frame.DataFrame, ts: TSDataset, segments: Optional[List[str]] = None, columns_num: int = 2, history_len: Union[int, Literal['all']] = 0, figsize: Tuple[int, int] = (10, 5))[source]

Plot targets and forecast for backtest pipeline.

This function doesn’t support intersecting folds.

Parameters

  • forecast_df (pandas.core.frame.DataFrame) – forecasted dataframe with timeseries data

  • ts (TSDataset) – dataframe of timeseries that was used for backtest

  • segments (Optional[List[str]]) – segments to plot

  • columns_num (int) – number of subplots columns

  • history_len (Union[int, Literal['all']]) – length of pre-backtest history to plot, if value is “all” then plot all the history

  • figsize (Tuple[int, int]) – size of the figure per subplot with one segment in inches

Raises

  • ValueError: – if history_len is negative

  • ValueError: – if folds are intersecting

plot_backtest_interactive(forecast_df: pandas.core.frame.DataFrame, ts: TSDataset, segments: Optional[List[str]] = None, history_len: Union[int, Literal['all']] = 0, figsize: Tuple[int, int] = (900, 600)) → plotly.graph_objs._figure.Figure[source]

Plot targets and forecast for backtest pipeline using plotly.

Parameters

  • forecast_df (pandas.core.frame.DataFrame) – forecasted dataframe with timeseries data

  • ts (TSDataset) – dataframe of timeseries that was used for backtest

  • segments (Optional[List[str]]) – segments to plot

  • history_len (Union[int, Literal['all']]) – length of pre-backtest history to plot, if value is “all” then plot all the history

  • figsize (Tuple[int, int]) – size of the figure in pixels

Returns

result of plotting

Return type

go.Figure

Raises

  • ValueError: – if history_len is negative

  • ValueError: – if folds are intersecting

plot_forecast(forecast_ts: Union[TSDataset, List[TSDataset], Dict[str, TSDataset]], test_ts: Optional[TSDataset] = None, train_ts: Optional[TSDataset] = None, segments: Optional[List[str]] = None, n_train_samples: Optional[int] = None, columns_num: int = 2, figsize: Tuple[int, int] = (10, 5), prediction_intervals: bool = False, quantiles: Optional[List[float]] = None)[source]

Plot of prediction for forecast pipeline.

Parameters

  • forecast_ts (Union[TSDataset, List[TSDataset], Dict[str, TSDataset]]) –

    there are several options:

    1. Forecasted TSDataset with timeseries data, single-forecast mode

    2. List of forecasted TSDatasets, multi-forecast mode

    3. Dictionary with forecasted TSDatasets, multi-forecast mode

  • test_ts (Optional[TSDataset]) – TSDataset with timeseries data

  • train_ts (Optional[TSDataset]) – TSDataset with timeseries data

  • segments (Optional[List[str]]) – segments to plot; if not given plot all the segments from forecast_df

  • n_train_samples (Optional[int]) – length of history of train to plot

  • columns_num (int) – number of graphics columns

  • figsize (Tuple[int, int]) – size of the figure per subplot with one segment in inches

  • prediction_intervals (bool) – if True prediction intervals will be drawn

  • quantiles (Optional[List[float]]) – List of quantiles to draw, if isn’t set then quantiles from a given dataset will be used. In multi-forecast mode, only quantiles present in each forecast will be used.

Raises

ValueError: – if the format of forecast_ts is unknown

plot_forecast_decomposition(forecast_ts: TSDataset, test_ts: Optional[TSDataset] = None, mode: Union[Literal['per-component'], Literal['joint']] = 'per-component', segments: Optional[List[str]] = None, columns_num: int = 1, figsize: Tuple[int, int] = (10, 5), show_grid: bool = False)[source]

Plot of prediction and its components.

Parameters

  • forecast_ts (TSDataset) – forecasted TSDataset with timeseries data, single-forecast mode

  • test_ts (Optional[TSDataset]) – TSDataset with timeseries data

  • mode (Union[Literal['per-component'], typing.Literal['joint']]) –

    Components plotting type

    1. per-component – plot each component in separate axes

    2. joint – plot all the components in the same axis

  • segments (Optional[List[str]]) – segments to plot; if not given plot all the segments

  • columns_num (int) – number of graphics columns; when mode=``per-component`` all plots will be in the single column

  • figsize (Tuple[int, int]) – size of the figure per subplot with one segment in inches

  • show_grid (bool) – whether to show grid for each chart

Raises

  • ValueError: – if components aren’t present in forecast_ts

  • NotImplementedError: – unknown mode is given

plot_metric_per_segment(metrics_df: pandas.core.frame.DataFrame, metric_name: str, ascending: bool = False, per_fold_aggregation_mode: str = 'mean', top_k: Optional[int] = None, barplot_params: Optional[Dict[str, Any]] = None, figsize: Tuple[int, int] = (10, 5))[source]

Plot barplot with per-segment metrics.

Parameters

  • metrics_df (pandas.core.frame.DataFrame) – dataframe with metrics calculated on the backtest

  • metric_name (str) – name of the metric to visualize

  • ascending (bool) –

    • If True, small values at the top;

    • If False, big values at the top.

  • per_fold_aggregation_mode (str) – how to aggregate metrics over the folds if they aren’t already aggregated (see PerFoldAggregation)

  • top_k (Optional[int]) – number segments to show after ordering according to ascending

  • barplot_params (Optional[Dict[str, Any]]) – dictionary with parameters for plotting, seaborn.barplot() is used

  • figsize (Tuple[int, int]) – size of the figure per subplot with one segment in inches

Raises

  • ValueError: – if metric_name isn’t present in metrics_df

  • NotImplementedError: – unknown per_fold_aggregation_mode is given

plot_residuals(forecast_df: pandas.core.frame.DataFrame, ts: TSDataset, feature: Union[str, Literal['timestamp']] = 'timestamp', transforms: Sequence[Transform] = (), segments: Optional[List[str]] = None, columns_num: int = 2, figsize: Tuple[int, int] = (10, 5))[source]

Plot residuals for predictions from backtest against some feature.

Parameters

  • forecast_df (pandas.core.frame.DataFrame) – forecasted dataframe with timeseries data

  • ts (TSDataset) – dataframe of timeseries that was used for backtest

  • feature (Union[str, Literal['timestamp']]) – feature name to draw against residuals, if “timestamp” plot residuals against the timestamp

  • transforms (Sequence[Transform]) – sequence of transforms to get feature column

  • segments (Optional[List[str]]) – segments to use

  • columns_num (int) – number of columns in subplots

  • figsize (Tuple[int, int]) – size of the figure per subplot with one segment in inches

Raises

ValueError: – if feature isn’t present in the dataset after applying transformations

Notes

Parameter transforms is necessary because some pipelines doesn’t save features in their forecasts, e.g. etna.ensembles pipelines.

prediction_actual_scatter_plot(forecast_df: pandas.core.frame.DataFrame, ts: TSDataset, segments: Optional[List[str]] = None, columns_num: int = 2, figsize: Tuple[int, int] = (10, 5))[source]

Plot scatter plot with forecasted/actual values for segments.

Parameters

  • forecast_df (pandas.core.frame.DataFrame) – forecasted dataframe with timeseries data

  • ts (TSDataset) – dataframe of timeseries that was used for backtest

  • segments (Optional[List[str]]) – segments to plot

  • columns_num (int) – number of columns in subplots

  • figsize (Tuple[int, int]) – size of the figure per subplot with one segment in inches

qq_plot(residuals_ts: TSDataset, qq_plot_params: Optional[Dict[str, Any]] = None, segments: Optional[List[str]] = None, columns_num: int = 2, figsize: Tuple[int, int] = (10, 5))[source]

Plot Q-Q plots for segments.

Parameters

  • residuals_ts (TSDataset) – dataset with the time series, expected to be the residuals of the model

  • qq_plot_params (Optional[Dict[str, Any]]) – dictionary with parameters for qq plot, statsmodels.graphics.gofplots.qqplot() is used

  • segments (Optional[List[str]]) – segments to plot

  • columns_num (int) – number of columns in subplots

  • figsize (Tuple[int, int]) – size of the figure per subplot with one segment in inches