[BUG] RandomForestRegressor to_json functionality broken

[jelc53]

  File "<string>", line 1, in <module>
  File "/virtualenvs/smartshift-load-forecasting-9TtSrW0h-py3.9/lib/python3.9/site-packages/sklearn_json/__init__.py", line 122, in to_json
    json.dump(serialize_model(model), model_json)
  File "/virtualenvs/smartshift-load-forecasting-9TtSrW0h-py3.9/lib/python3.9/site-packages/sklearn_json/__init__.py", line 57, in serialize_model
    return reg.serialize_random_forest_regressor(model)
  File "/virtualenvs/smartshift-load-forecasting-9TtSrW0h-py3.9/lib/python3.9/site-packages/sklearn_json/regression.py", line 307, in serialize_random_forest_regressor
    'min_impurity_split': model.min_impurity_split,
AttributeError: 'RandomForestRegressor' object has no attribute 'min_impurity_split'```

[shon-otmazgin]

I have t same issue. Did you find any workaround?

[jamesheaton]

Also getting the same error, I've not found any workarounds

[ZumelzuR]

as you not use min_impurity_split in last versions of sklearn, you should comment on the code of the library.

Also you should do replace n_features by n_features_in_, and check the oficial documentation to check which values work for your version.

My changes was

def deserialize_random_forest(model_dict):
    model = RandomForestClassifier(**model_dict['params'])
    estimators = [deserialize_decision_tree(decision_tree) for decision_tree in model_dict['estimators_']]
    model.estimators_ = np.array(estimators)

    model.classes_ = np.array(model_dict['classes_'])
    model.n_features_ = model_dict['n_features_']
    model.n_outputs_ = model_dict['n_outputs_']
    model.max_depth = model_dict['max_depth']
    model.min_samples_split = model_dict['min_samples_split']
    model.min_samples_leaf = model_dict['min_samples_leaf']
    model.min_weight_fraction_leaf = model_dict['min_weight_fraction_leaf']
    model.max_features = model_dict['max_features']
    model.max_leaf_nodes = model_dict['max_leaf_nodes']
    model.min_impurity_decrease = model_dict['min_impurity_decrease']
    # model.min_impurity_split = model_dict['min_impurity_split']

and

def serialize_random_forest(model):
    serialized_model = {
        'meta': 'rf',
        'max_depth': model.max_depth,
        'min_samples_split': model.min_samples_split,
        'min_samples_leaf': model.min_samples_leaf,
        'min_weight_fraction_leaf': model.min_weight_fraction_leaf,
        'max_features': model.max_features,
        'max_leaf_nodes': model.max_leaf_nodes,
        'min_impurity_decrease': model.min_impurity_decrease,
        # 'min_impurity_split': model.min_impurity_split,
        'n_features_in_': model.n_features_in_,
        'n_features': model.n_features_in_,
        'n_outputs_': model.n_outputs_,
        'classes_': model.classes_.tolist(),
        'estimators_': [serialize_decision_tree(decision_tree) for decision_tree in model.estimators_],
        'params': model.get_params()
    }

and

def serialize_decision_tree(model):
    print(model)
    tree, dtypes = serialize_tree(model.tree_)
    serialized_model = {
        'meta': 'decision-tree',
        'feature_importances_': model.feature_importances_.tolist(),
        'max_features_': model.max_features_,
        'n_classes_': int(model.n_classes_),
        'n_features_in_': model.n_features_in_,
        'n_features_': model.n_features_in_,
        'n_outputs_': model.n_outputs_,
        'tree_': tree,
        'classes_': model.classes_.tolist(),
        'params': model.get_params()
    }

If I have some time and this work good for me I will do a PR