Fix analysis.forecast.plots.metric_per_segment_distribution_plot to handle None from metrics

[d-a-bunin]

Before submitting (must do checklist)

• Did you read the contribution guide?
• Did you update the docs? We use Numpy format for all the methods and classes.
• Did you write any new necessary tests?
• Did you update the CHANGELOG?

Proposed Changes

See #533.

Closing issues

Closes #533.

[github-actions]

🚀 Deployed on https://deploy-preview-543--etna-docs.netlify.app

[d-a-bunin]

Example of plotting:

import pandas as pd
import matplotlib.pyplot as plt
from loguru import logger

from etna.analysis import metric_per_segment_distribution_plot


def get_metrics_df_with_folds() -> pd.DataFrame:
    df = pd.DataFrame(
        {
            "segment": ["segment_0"] * 3 + ["segment_1"] * 3 + ["segment_2"] * 3,
            "MAE": [1.0, 2.0, 3.0, 2.0, 3.0, 4.0, 3.0, 4.0, 5.0],
            "MSE": [None, 3.0, 4.0, 3.0, 4.0, 5.0, 5.0, 6.0, 7.0],
            "MAPE": [None, None, None, 20.0, 30.0, 40.0, 30.0, 40.0, 50.0],
            "SMAPE": [None, 20.0, 30.0, None, 50.0, 60.0, None, 70.0, 80.0],
            "RMSE": [None, None, None, None, None, None, None, None, None],
            "fold_number": [0, 1, 2, 0, 1, 2, 0, 1, 2],
        }
    )
    return df


def main():
    metrics_df = get_metrics_df_with_folds()

    metric_per_segment_distribution_plot(metrics_df=metrics_df, metric_name="MAPE", per_fold_aggregation_mode=None, plot_type="hist")
    plt.savefig("mape_none_hist.png", bbox_inches="tight")

    metric_per_segment_distribution_plot(metrics_df=metrics_df, metric_name="SMAPE", per_fold_aggregation_mode=None, plot_type="hist")
    plt.savefig("smape_none_hist.png", bbox_inches="tight")

    metric_per_segment_distribution_plot(metrics_df=metrics_df, metric_name="SMAPE", per_fold_aggregation_mode="mean", plot_type="hist")
    plt.savefig("smape_mean_hist.png", bbox_inches="tight")

    try:
        metric_per_segment_distribution_plot(metrics_df=metrics_df, metric_name="RMSE", per_fold_aggregation_mode=None, plot_type="hist")
        plt.savefig("rmse_none_hist.png", bbox_inches="tight")
    except ValueError:
        logger.info("No success with RMSE!")


if __name__ == "__main__":
    main()

mape_none_hist:

smape_none_hist

smape_mean_hist:

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 90.52%. Comparing base (7b37537) to head (bd87574).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #543      +/-   ##
==========================================
+ Coverage   90.48%   90.52%   +0.03%     
==========================================
  Files         256      256              
  Lines       17373    17375       +2     
==========================================
+ Hits        15720    15728       +8     
+ Misses       1653     1647       -6     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[d-a-bunin]

Example without intersection:

import pandas as pd
import matplotlib.pyplot as plt
from loguru import logger

from etna.analysis import metric_per_segment_distribution_plot


def get_metrics_df_with_folds() -> pd.DataFrame:
    df = pd.DataFrame(
        {
            "segment": ["segment_0"] * 3 + ["segment_1"] * 3 + ["segment_2"] * 3,
            "MAE": [1.0, 2.0, 3.0, 2.0, 3.0, 4.0, 3.0, 4.0, 5.0],
            "MSE": [None, 3.0, 4.0, 3.0, 4.0, 5.0, 5.0, 6.0, 7.0],
            "MAPE": [None, None, None, 20.0, 50.0, 90.0, 30.0, 60.0, 100.0],
            "SMAPE": [None, 20.0, 60.0, None, 30.0, 70.0, None, 40.0, 80.0],
            "RMSE": [None, None, None, None, None, None, None, None, None],
            "fold_number": [0, 1, 2, 0, 1, 2, 0, 1, 2],
        }
    )
    return df


def main():
    metrics_df = get_metrics_df_with_folds()

    metric_per_segment_distribution_plot(metrics_df=metrics_df, metric_name="MAPE", per_fold_aggregation_mode=None, plot_type="hist")
    plt.savefig("mape_none_hist.png", bbox_inches="tight")

    metric_per_segment_distribution_plot(metrics_df=metrics_df, metric_name="SMAPE", per_fold_aggregation_mode=None, plot_type="hist")
    plt.savefig("smape_none_hist.png", bbox_inches="tight")

    metric_per_segment_distribution_plot(metrics_df=metrics_df, metric_name="SMAPE", per_fold_aggregation_mode="mean", plot_type="hist")
    plt.savefig("smape_mean_hist.png", bbox_inches="tight")

    try:
        metric_per_segment_distribution_plot(metrics_df=metrics_df, metric_name="RMSE", per_fold_aggregation_mode=None, plot_type="hist")
        plt.savefig("rmse_none_hist.png", bbox_inches="tight")
    except ValueError:
        logger.info("No success with RMSE!")


if __name__ == "__main__":
    main()

mape_none_hist

smape_none_hist

smape_mean_hist