WindowedMetricPlotter#

class openstef_beam.analysis.plots.windowed_metric_plotter.WindowedMetricPlotter[source]

Bases: object

Creates time series plots of performance metrics using windowed analysis.

This plotter visualizes how model performance changes over time by computing metrics within rolling time windows. This helps identify temporal patterns, seasonal variations, and long-term trends in forecast accuracy.

The plots show:

  • Performance metrics over time for each model

  • Rolling averages to smooth out short-term fluctuations

  • Comparative analysis across multiple models

  • Trend identification for performance monitoring

Example

Basic usage for RMSE over time

>>> from datetime import datetime
>>> plotter = WindowedMetricPlotter()
>>> timestamps = [datetime(2024, 1, 1), datetime(2024, 1, 2)]
>>> rmse_values = [0.12, 0.15]
>>> _ = plotter.add_model("XGBoost", timestamps, rmse_values)
>>> _ = plotter.set_window_size("1D")
>>> fig = plotter.plot(title="RMSE Over Time", metric_name="RMSE")
>>> type(fig).__name__
'Figure'
__init__()[source]

Initialize the plotter with empty data containers.

add_model(model_name: str, timestamps: list[datetime], metric_values: list[float]) → WindowedMetricPlotter[source]

Add a model’s metric values and timestamps to the plot.

Parameters:

  • model_name (str) – The name of the model.

  • timestamps (list[datetime]) – List of datetime objects for the x-axis.

  • metric_values (list[float]) – List of metric values for the y-axis.

  • model_name

  • timestamps

  • metric_values

Returns:

The current instance for method chaining.

Return type:

WindowedMetricPlotter

Raises:

ValueError – If timestamps and metric values have different lengths.

Return type:

WindowedMetricPlotter

set_window_size(window_size: str) → WindowedMetricPlotter[source]

Set the window size used for metric calculation.

Parameters:

  • window_size (str) – Description of the window size (e.g., “7 days”).

  • window_size

Returns:

The current instance for method chaining.

Return type:

WindowedMetricPlotter

Return type:

WindowedMetricPlotter

plot(title: str = 'Metric over time', metric_name: str = 'Metric') → Figure[source]

Create and return a line chart of metrics over time for all added models.

Parameters:

  • title (str) – Main title of the plot.

  • metric_name (str) – Name of the metric being plotted.

  • title

  • metric_name

Returns:

The resulting plot.

Return type:

plotly.graph_objects.Figure

Raises:

ValueError – If no model data has been added.

Return type:

Figure