Answer-first summary for fast verification
Answer: Implement a continuous monitoring system for data drift and concept drift, with automated alerts and retraining processes triggered when significant deviations are detected.
Implementing a continuous monitoring system for data drift and concept drift is the most effective strategy because it proactively addresses the root cause of the performance drop—changes in input data distribution and customer behavior. Automated alerts and retraining processes ensure the model adapts to these changes in real-time, maintaining high accuracy without requiring constant manual intervention. This approach is both cost-effective and scalable, aligning with the company's requirements for minimal manual intervention and cost. Feature selection or engineering alone does not address the dynamic nature of data drift, and periodic retraining without monitoring may not catch shifts in data distribution promptly.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
You are a Machine Learning Engineer at a retail company that uses a DNN regression model to predict customer spending. The model was deployed six months ago and initially showed high accuracy. However, recent evaluations indicate a significant drop in performance due to shifts in customer behavior and input data distribution caused by seasonal changes and a new marketing campaign. The company requires a solution that not only addresses the current performance drop but also prevents future discrepancies with minimal manual intervention and cost. Which of the following strategies is the MOST effective to achieve this goal? (Choose one)
A
Conduct feature selection to reduce the number of features and retrain the model monthly with these selected features, assuming that fewer features will make the model more robust to changes.
B
Implement a continuous monitoring system for data drift and concept drift, with automated alerts and retraining processes triggered when significant deviations are detected.
C
Retrain the model using the latest data, employing a hyperparameter tuning service to optimize the L2 regularization parameter, focusing on improving model complexity.
D
Perform feature engineering to create more relevant features based on the current data distribution and retrain the model with this enhanced feature set.
No comments yet.