Hands-On Exercise: Build a Defect Prediction System CT-AI Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A team is starting a defect prediction initiative for a new microservices project. They have access to CI/CD logs, container metrics, and code repositories. Which data source is most valuable for predicting defects at the service level?

   • ADeveloper names from commit history.
   • BContainer CPU usage metrics.
   • CCI/CD logs that record build failures and test results per service.
   • DNumber of microservices in the project.
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A team has a defect dataset with 10,000 modules, but only 500 are defective. They have tried SMOTE to balance the data, but the model still shows poor recall. What is the most likely reason?

   • AThe model is overfitting the synthetic samples.
   • BSMOTE should only be applied to the test set, not the training set.
   • CThe dataset is too large for SMOTE to handle.
   • DThe synthetic samples generated by SMOTE may not represent the true distribution of defective modules.
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A team has trained a defect prediction model and evaluated it on a holdout test set. The model achieves 90% accuracy, but the recall for defective modules is only 40%. The dataset is imbalanced with 10% defective modules. What is the most likely reason for this discrepancy?

   • AThe model is underfitting the training data.
   • BThe model is overfitting the training data.
   • CThe test set is too small to be reliable.
   • DThe model is biased toward the majority class (non-defective modules).
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   Which metric is most appropriate to evaluate a defect prediction model when the cost of missing a defective module is high?

   • AAccuracy
   • BF1-score
   • CPrecision
   • DRecall
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   What type of machine learning problem is defect prediction typically framed as?

   • AA reinforcement learning problem, learning from rewards and penalties
   • BA clustering problem, grouping similar modules together
   • CA classification problem, predicting whether a module is defective or not
   • DA regression problem, predicting the number of defects in a module
   Reveal Solution Discussion Select an answer first