Saving and loading MetricFrame

In this section, we will discuss how to save MetricFrame in pickle format and how to load it from a stored pickle file.

>>> y_true = [0, 1, 1, 1, 1, 0, 1, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1]
>>> y_pred = [0, 0, 1, 0, 1, 1, 1, 0, 0, 1, 1, 1, 1, 0, 0, 1, 1, 0]
>>> sf_data = ['b', 'b', 'a', 'b', 'b', 'c', 'c', 'c', 'a',
...            'a', 'c', 'a', 'b', 'c', 'c', 'b', 'c', 'c']
>>> from fairlearn.metrics import MetricFrame
>>> from sklearn.metrics import accuracy_score
>>> import pickle
>>> metric_frame = MetricFrame(metrics=accuracy_score,
...                            y_true=y_true,
...                            y_pred=y_pred,
...                            sensitive_features=sf_data)
>>> metric_frame.overall.item()
0.444...
>>> metric_frame.by_group
sensitive_feature_0
a    0.250000
b    0.666667
c    0.375000
Name: accuracy_score, dtype: float64
>>> file_name = 'metric_frame_save_load_example.pkl'
>>> pickle.dump(metric_frame, open(file_name, 'wb'))
    >>> loaded_metric_frame = pickle.load(open(file_name, 'rb'))
>>> loaded_metric_frame.overall.item()
0.444...
>>> loaded_metric_frame.by_group
sensitive_feature_0
a    0.250000
b    0.666667
c    0.375000
Name: accuracy_score, dtype: float64