ML Workflow CT-AI Practice Questions (Page 6)

Questions 26–28

1. 26application · medium

   A team is training a support vector machine (SVM) for a binary classification problem. They notice the model is underfitting: both training and validation accuracy are low. Which action is most likely to improve the model's performance?

   • AUse a smaller training dataset to reduce overfitting
   • BIncrease the model's capacity by using a more complex kernel
   • CReduce the number of features to simplify the model
   • DIncrease the regularization parameter to penalize complexity
   Reveal Solution Discussion Select an answer first
2. 27application · medium

   A company has deployed a model to recommend products to users. They want to ensure the model's recommendations remain relevant as user preferences change. What is the most appropriate approach?

   • AIncrease the model's complexity to predict future preferences
   • BKeep the model static to ensure consistent recommendations
   • CSet up a feedback loop to collect user interactions and periodically retrain the model
   • DReduce the number of recommendations to avoid user fatigue
   Reveal Solution Discussion Select an answer first
3. 28foundation · easy

   What is a key activity in monitoring a deployed ML model?

   • ASelecting the initial algorithm for the model
   • BSplitting the dataset into training and test sets
   • CCollecting new training data from scratch
   • DTracking performance metrics over time to detect drift
   Reveal Solution Discussion Select an answer first