Determine If Data Meets Solution Needs PMI-CPMAI Practice Questions (Page 5)

Questions 21–24

1. 21application · medium

   A healthcare startup is developing an AI model to detect skin lesions from images. They have 10,000 images, but 80% are of benign lesions and only 20% are malignant. The model will be used in a clinical setting where missing a malignant lesion is critical. The team is considering whether to proceed with training. What should they do?

   • ACollect more images of benign lesions to balance the dataset.
   • BAssess the sufficiency of the data for the clinical use case, considering both volume and class balance, before making a go/no-go decision.
   • CProceed with training because 10,000 images is a large dataset for image classification.
   • DProceed with training but use class weights to handle the imbalance.
   Reveal Solution Discussion Select an answer first
2. 22application · medium

   A streaming service is building a recommendation model. The data science team has access to viewing history, but the data is from the past six months. The business wants the model to capture long-term viewing trends. The team is assessing data sufficiency. What should they do?

   • AAssess whether the six-month period is sufficient to capture the long-term trends required by the business.
   • BCollect more data from the past year before training.
   • CUse the six months of data and rely on the model to extrapolate long-term trends.
   • DProceed with the six months of data because it is recent and relevant.
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   After a comprehensive data readiness assessment, the team finds that the data is insufficient in volume and has significant quality issues. What is the most appropriate go/no-go decision?

   • AGo, because the team can proceed and fix issues later
   • BGo, but only for a limited pilot
   • CNo-go, and develop a plan to address the data deficiencies
   • DNo-go, and abandon the project entirely
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   A healthcare provider is building a model to predict patient readmission. The data science team has electronic health records (EHR) from one hospital. The model will be deployed across multiple hospitals with different patient populations. The team must make a go/no-go decision. What is the most critical factor to consider?

   • AThe data includes a sufficient number of readmission cases to train the model.
   • BThe data is recent and reflects current medical practices.
   • CThe data is representative of the patient populations at all deployment hospitals.
   • DThe total number of patient records is large enough to train a deep learning model.
   Reveal Solution Discussion Select an answer first