Hands-On Exercise: Build a Defect Prediction System CT-AI Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A defect prediction model is being evaluated for a safety-critical system. The cost of missing a defect is very high, but the team also wants to avoid overwhelming testers with too many false alarms. They have two models: Model A has precision 0.8 and recall 0.6; Model B has precision 0.6 and recall 0.9. Which model should they choose?

   • AModel B, because higher recall reduces the risk of missing defects, which is critical in safety-critical systems.
   • BChoose the model with the higher F1-score, regardless of the context.
   • CNeither model is suitable; they need a model with both high precision and high recall.
   • DModel A, because higher precision reduces false alarms and is more efficient.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   What type of data is essential for training a supervised defect prediction model?

   • AUser feedback from the production environment
   • BHistorical defect data that labels modules as defective or not
   • CTest execution logs from the current sprint
   • DOnly the source code of the current release
   Reveal Solution Discussion Select an answer first
4. 18expert · hard

   A defect prediction model is used to prioritize testing. The model predicts that a certain module has a 90% probability of containing a defect. However, manual testing of that module reveals no defects. What is the most appropriate interpretation?

   • AThe model is broken and should be retrained immediately.
   • BThe model may have a false positive, but it is still useful for prioritization; the module should still be tested thoroughly.
   • CThe module is definitely defect-free and can be skipped in future testing.
   • DThe model's predictions should be ignored because they are not always correct.
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A defect prediction model is being trained on a dataset with 10,000 modules and 50 features. The team wants to use a classification algorithm that can capture non-linear relationships and provide feature importance. Which algorithm is most suitable?

   • ALinear Regression
   • BApriori Algorithm
   • CK-Means Clustering
   • DRandom Forest
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   What is the purpose of normalizing features in a defect prediction dataset?

   • ATo increase the number of features in the dataset
   • BTo convert all features into binary values
   • CTo bring all features to a similar scale so that no single feature dominates the model due to its magnitude
   • DTo remove features that are not correlated with defects
   Reveal Solution Discussion Select an answer first