Databricks Machine Learning Professional Question 275
Single answerA machine learning model predicting customer churn in a subscription-based service has been deployed into production. Over the last three months, you observe a consistent drop in its accuracy, and data analysis shows a significant change in customer behavior patterns during that time. What is the most appropriate course of action to address this issue?
- A
Retrain the model with recent data that reflects the changed customer behavior and redeploy it.
- B
Increase the frequency of model inference to handle the changes in real-time.
- C
Add more features to the existing dataset to improve the model's accuracy.
- D
Perform hyperparameter tuning on the existing deployed model to optimize its performance.
Show answer and explanation
Correct answer: A
Explanation
Data drift refers to changes in the statistical properties of input data over time, which can degrade a model's performance. In this scenario, the model's accuracy has dropped due to significant changes in customer behavior, indicating data drift. Retraining the model with updated data ensures it learns from the new patterns and remains effective in predicting customer churn.
- A. Correct.
Retraining the model with recent data that reflects the changed customer behavior is the correct solution because the significant change in behavior patterns is likely causing data drift. Retraining ensures the model adapts to the new distribution of data.
- B. Incorrect.
Increasing the frequency of model inference does not address the underlying issue of data drift and will not improve the model's accuracy in this situation.
- C. Incorrect.
Adding more features to the dataset might be helpful in some cases, but it does not directly address the issue of data drift caused by changes in customer behavior.
- D. Incorrect.
Hyperparameter tuning optimizes the model's performance but does not solve the problem of drift caused by changes in the underlying data distribution.