Databricks Certified Machine Learning Professional — Question 69
A machine learning engineer is in the process of implementing a feature drift monitoring solution. They are planning to use the following steps:
1. Measure the distributions of each feature variable in the training set
2. Deploy a model to production
3. Measure the distributions of each feature variable in inference
4. _______
Which action should be completed as Step #4?
Answer options
- A. Retrain the model based on any new feature variables that have been added to the feature set
- B. Obtain the observed values (actual) feature values and compare to the predicted values
- C. Measure the latency of the model's prediction time
- D. Run a statistical test to determine if there are changes in the feature variable distribution over time
Correct answer: D
Explanation
The correct answer is D because running a statistical test helps to identify if there are significant changes in the feature variable distributions, which is critical for monitoring feature drift. Options A and B are related to model performance and retraining, but they do not directly address the monitoring of feature drift. Option C is focused on performance measurement rather than distribution analysis.