AI-Based System Testing Fundamentals CT-AI Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   Which scenario would most likely require a more rigorous and extensive test approach for an ML system?

   • AA system that predicts weather for a general forecast
   • BA system that diagnoses medical conditions from patient data
   • CA system that recommends movies based on viewing history
   • DA system that sorts emails into spam or inbox
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A team is testing an image recognition model for a self-driving car. They have a large dataset of daytime images but very few nighttime or rainy images. What is the best way to prepare test data to evaluate the model's robustness?

   • ASplit the existing dataset into training and test sets, ensuring a balanced distribution of conditions.
   • BTest only on real-world nighttime and rainy images collected from public datasets.
   • CUse only the available daytime images to test the model, as they represent the majority of driving conditions.
   • DAugment the test set with synthetic variations (e.g., brightness, contrast, rain effects) to simulate diverse conditions.
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   Why is it important to use a separate test dataset that was not used during training?

   • ATo increase the size of the training dataset
   • BTo evaluate the model's performance on unseen data
   • CTo ensure the model is trained on more diverse data
   • DTo reduce the time needed for training
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   An AI system for predicting patient readmission uses a dataset where the 'insurance' feature has many missing values. The team decides to remove all records with missing insurance data. What is a potential consequence?

   • AThe test data will be more representative of the training data.
   • BThe test set may become biased if the missingness is not random.
   • CThe model will be more robust to missing data.
   • DThe model will become more accurate.
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A team is testing an AI system that predicts hospital readmission risk. The training data was collected from a single hospital and contains many missing values for patients over 65. The test data is from the same hospital but covers a different time period. What is the most important data quality issue to address before testing?

   • AThe training data is from a single hospital, so the model will not generalize.
   • BThe training data has a high proportion of missing values for elderly patients, which could introduce bias.
   • CThe test data is too small to provide statistically significant results.
   • DThe test data is not representative of the training data distribution.
   Reveal Solution Discussion Select an answer first