Should we use `get_feature_names_out` instead of `get_feature_names`?

[gaugup]

With a newer version of scikit-learn 1.2.0 we got the following error emanating from _causal_analysis.py:-

{
  "code": "ExecutionFailed",
  "category": "UserError",
  "message": {
    "NonCompliant": "Process 'python' exited with code 1 and error message 'Execution failed. Process exited with status code 1. Error: Traceback (most recent call last):\n  File \"create_causal.py\", line 121, in <module>\n    main(args)\n  File \"create_causal.py\", line 98, in main\n    rai_i.compute()\n  File \"/azureml-envs/azureml_6de821483c371657b39f098361948b90/lib/python3.8/site-packages/responsibleai/rai_insights/rai_insights.py\", line 389, in compute\n    manager.compute()\n  File \"/azureml-envs/azureml_6de821483c371657b39f098361948b90/lib/python3.8/site-packages/responsibleai/managers/causal_manager.py\", line 341, in compute\n    policy = self._create_policy(\n  File \"/azureml-envs/azureml_6de821483c371657b39f098361948b90/lib/python3.8/site-packages/responsibleai/managers/causal_manager.py\", line 244, in _create_policy\n    tree = causal_analysis._policy_tree_output(\n  File \"/azureml-envs/azureml_6de821483c371657b39f098361948b90/lib/python3.8/site-packages/econml/solutions/causal_analysis/_causal_analysis.py\", line 1520, in _policy_tree_output\n    (policy_val, always_trt)) = self._tree(True, Xtest, feature_index,\n  File \"/azureml-envs/azureml_6de821483c371657b39f098361948b90/lib/python3.8/site-packages/econml/solutions/causal_analysis/_causal_analysis.py\", line 1441, in _tree\n    return intrp, result.X_transformer.get_feature_names(self.feature_names_), treatment_names, policy_values\n  File \"/azureml-envs/azureml_6de821483c371657b39f098361948b90/lib/python3.8/site-packages/econml/solutions/causal_analysis/_causal_analysis.py\", line 228, in get_feature_names\n    cats = self.one_hot_encoder.get_feature_names(\nAttributeError: 'OneHotEncoder' object has no attribute 'get_feature_names'\n'. Please check the log file 'user_logs/std_log.txt' for more details."
  },
  "details": [
    {
      "name": "exit_codes",
      "value": {
        "Literal": {
          "Compliant": "1"
        }
      }
    }
  ],
  "error": null,
  "node_info": null
}

Looks like the method get_feature_names is deprecated in newer version of scikit-learn. Is it possible to change to API get_feature_names_out?

[kbattocchi]

Thanks for pointing this out; we should support both so that the library can be used with newer or older versions of sklearn. I'll have a PR for this shortly.