Dealing with a MissingMetadataException

The problem

Running the fit, calculate or estimate methods of a Calculator or Estimator fails by returning a MissingMetadataException.

nannyml.exceptions.MissingMetadataException: metadata is still missing values for ['predicted_probability_column_name'].

The solution

The MissingMetadataException is raised when the model metadata used to create the Calculator or Estimator is not complete, i.e. it is missing some required properties.

The exception will list the properties it is missing, as shown in the problem statement.

Assume md is the model metadata object used, predicted_probability_column_name is the property missing and in your data the predicted probabilities are located in the model_probas column.

The following snippet should help you prevent the exception by completing the metadata manually:

>>> md.is_complete()  # just checking
(False, ['predicted_probability_column_name'])
>>> md.predicted_probability_column_name = 'model_probas'
>>> md.is_complete()
(True, [])

Any metadata property can be set or updated.

Related reads

To read more on metadata, read the tutorial on providing metadata.

NannyML can automatically extract metadata from a data sample if it follows some naming conventions.