AI-900 Question 128
Select 3A data scientist wants to build a machine learning model to predict customer churn without manually selecting algorithms or tuning hyperparameters. How can Azure Machine Learning's automated machine learning capabilities assist in this scenario?
- A
It automatically selects the best machine learning algorithm for the dataset.
- B
It tunes hyperparameters to optimize model performance.
- C
It requires the data scientist to write custom Python code for feature engineering.
- D
It generates explainability reports to help understand the model's predictions.
- E
It allows the data scientist to manually test each algorithm to find the best one.
Show answer and explanation
Correct answers: A, B, D
Explanation
Azure Automated Machine Learning enables data scientists to automate key steps in the machine learning workflow, such as algorithm selection, hyperparameter tuning, and model explainability. This allows users to focus on higher-level tasks while ensuring optimal model performance without manual intervention. However, tasks like manual algorithm testing and writing custom code for feature engineering are not necessary as they are handled automatically.
- A. Correct.
Azure Automated Machine Learning automatically evaluates multiple machine learning algorithms and selects the one that performs best for the given dataset. This reduces the need for manual algorithm selection.
- B. Correct.
Automated ML includes automated hyperparameter tuning to optimize the parameters of the chosen algorithm for better performance.
- C. Incorrect.
While automated ML can work with preprocessed data, it does not require the data scientist to write custom Python code for feature engineering. It can automatically perform some feature engineering tasks.
- D. Correct.
Azure Automated ML provides model explainability tools, such as feature importance charts, to help understand how the model makes predictions.
- E. Incorrect.
Automated ML is designed to automate the process of testing multiple algorithms, so manual testing of algorithms is not required.