Hands-On Exercise: Build a Defect Prediction System CT-AI Practice Questions (Page 1)

Questions 1–5

1. 1expert · hard

   A team is building a defect prediction model for a project that has just started, with only one release of data available. They want to predict defects in the next release. What is the most significant challenge they face?

   • ALimited historical data may lead to an unreliable model that does not generalize well.
   • BThe model will overfit the training data.
   • CFeature selection is impossible with so little data.
   • DThe data is likely to be imbalanced, but this can be easily fixed with SMOTE.
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   A team is training a defect prediction model using a dataset with 5,000 modules and 30 features. They want to use a model that is easy to interpret and can handle both numerical and categorical features. Which algorithm is most appropriate?

   • ADecision Tree
   • BSupport Vector Machine (SVM) with RBF kernel
   • CK-Nearest Neighbors (KNN)
   • DNeural Network
   Reveal Solution Discussion Select an answer first
4. 3expert · hard

   A defect prediction model is trained on data from a previous release and will be used to predict defects in the next release. The team evaluates the model using cross-validation on the historical data and gets an AUC-ROC of 0.85. However, when applied to the new release, the performance is much lower. What is the most likely cause?

   • AThe test set is too small.
   • BConcept drift: the relationship between features and defects may have changed between releases.
   • CThe model was not trained on enough data.
   • DThe model is overfitting the training data.
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A defect prediction model is being built for a new project with limited historical data. The team wants to create features that capture the likelihood of defects. Which feature engineering approach is most likely to be effective?

   • AUse the number of developers on the team as a feature.
   • BUse only the module name as a feature.
   • CCreate a feature that is the current date and time.
   • DCreate features based on code churn, such as the number of lines added and deleted per module.
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   How can the output of a defect prediction model be used to prioritize testing efforts?

   • ABy skipping testing on modules predicted to be defect-free
   • BBy focusing testing on modules predicted to have a high probability of defects
   • CBy replacing all manual testing with automated testing
   • DBy automatically fixing the defects in the predicted modules
   Reveal Solution Discussion Select an answer first