JSON Export of a scikit-learn Pipeline object

export_scikit_pipeline.py

import json

def fullname(o):
  return o.__module__ + "." + o.__class__.__name__

def export_pipeline(scikit_pipeline):
  """JSON export of a scikit-learn pipeline.
  
  Especially useful when paired with GridSearchCV, TPOT, etc.
    
  Example:

  best_model = GridSearchCV(
      some_pipeline,
      param_grid=some_tuning_parameters
  )
  best_model.fit(X=train_x, y=train_y)
  export_pipeline(best_model.best_estimator_)

  :param scikit_pipeline: a scikit-learn Pipeline object
  """
  steps_obj = {'steps':[]}
  for name, md in scikit_pipeline.steps:
      steps_obj['steps'].append({
          'name': name,
          'class_name': fullname(md),
          'params': md.get_params()
      })

  return steps_obj