AI Solution Development Methodologies and Lifecycle AAIA Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   Which AI development methodology is characterized by iterative cycles that combine data understanding, data preparation, modeling, and evaluation, and is specifically designed for data mining and analytics projects?

   • AWaterfall
   • BCRISP-DM
   • CAgile
   • DDevOps
   Reveal Solution Discussion Select an answer first
2. 22expert · hard

   A government agency is implementing an AI system to automate the processing of visa applications. The system will make recommendations to human officers, who have the final decision. The agency's union is concerned that the system will reduce the number of officers needed. The agency also faces a legal requirement to ensure that the system does not discriminate based on nationality. What is the most critical change management action to ensure successful adoption?

   • ADeploy the system in a pilot and demonstrate its efficiency gains to the union
   • BCommunicate to the union that the system will augment officers' work, not replace them, and involve them in the design
   • CProvide training to officers on how to use the system after deployment
   • DFocus on the technical development and ensure the model is non-discriminatory
   Reveal Solution Discussion Select an answer first
4. 23expert · hard

   A hospital is planning an AI system to predict patient deterioration. The system will use electronic health records (EHR) data. The hospital's legal team has identified that the data contains sensitive personal information and is subject to strict privacy regulations. The data science team has found that the EHR data is incomplete and inconsistent across departments. The hospital has a limited budget and cannot afford a large-scale data cleaning effort. What is the most critical feasibility assessment to perform before proceeding?

   • AData feasibility, including the cost to clean and standardize the EHR data and compliance with privacy regulations
   • BTechnical feasibility of integrating the AI system with the EHR system
   • CRegulatory feasibility of using EHR data for research purposes
   • DOrganizational feasibility of training clinicians to use the system
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   A credit card company is developing a model to detect fraudulent transactions. The dataset contains millions of transactions, but fraudulent transactions are extremely rare (0.01% of the data). The data science team has tried oversampling the minority class, but the model still performs poorly on new fraud cases. The team suspects that the oversampling is creating synthetic patterns that do not generalize. What should the team do to improve the model's ability to detect new fraud?

   • AReduce the number of features to simplify the model and avoid overfitting
   • BCollect more real fraud data from external sources or through collaboration with other financial institutions
   • CUse a more complex model, such as a deep neural network, to capture the patterns in the data
   • DUse a different resampling technique, such as SMOTE, to generate more realistic synthetic samples
   Reveal Solution Discussion Select an answer first
6. 25foundation · easy

   What is the primary purpose of monitoring model drift in a production AI system?

   • ATo ensure the model is compliant with regulations
   • BTo improve the model's accuracy on the training data
   • CTo detect when the model's performance degrades due to changes in data patterns
   • DTo reduce the latency of model predictions
   Reveal Solution Discussion Select an answer first