CERTIFIED-MACHINE-LEARNING-PROFESSIONAL Question #53: Real Exam Question with Answer & Explanation

Question

A data scientist has developed and logged a scikit-learn random forest model model, and then they ended their Spark session and terminated their cluster. After starting a new cluster, they want to review the feature_importances_ of the original model object. Which of the following lines of code can be used to restore the model object so that feature_importances_ is available?

Options

• Amlflow.load_model(model_uri)
• Bclient.list_artifacts(run_id)["feature-importances.csv"]
• Cmlflow.sklearn.load_model(model_uri)
• DThis can only be viewed in the MLflow Experiments UI
• Eclient.pyfunc.load_model(model_uri)

Explanation

Important note: The stated correct answer of A appears to contain an error - mlflow.load_model() is not a valid top-level MLflow function and would raise an AttributeError. The actual correct answer is C.

Option C (mlflow.sklearn.load_model(model_uri)) is correct because it uses the sklearn-flavored loader, which returns the native scikit-learn object - in this case, the actual RandomForestClassifier instance. Since feature_importances_ is a native sklearn attribute, it is fully accessible on the returned object.

Why the distractors are wrong:

• A (mlflow.load_model) - This function does not exist at the top level of the MLflow API; it would raise AttributeError.
• B (client.list_artifacts(...)["feature-importances.csv"]) - list_artifacts returns a list of FileInfo objects, not a dict, and you cannot index it like a dictionary; also, feature_importances_ is not automatically saved as a CSV artifact.
• D - The MLflow UI shows metrics and parameters, not reconstructed Python model objects with live attributes.
• E (client.pyfunc.load_model) - MlflowClient has no pyfunc attribute; the correct call would be mlflow.pyfunc.load_model(), and even that returns a generic PyFuncModel wrapper that does not expose sklearn-specific attributes like feature_importances_.

Memory tip: Think of it as speaking the model's language - use the flavor-specific loader (mlflow.sklearn, mlflow.xgboost, etc.) when you need the native object and its native attributes. pyfunc gives you a generic wrapper for inference only.

No community discussion yet for this question.