Objective 2.5 Address Business Risks, Ethical Concerns, and Related Concepts in Data Exploration/feature Engineering AIP-210 Practice Questions (Page 6)

Questions 26–30

1. 26foundation · easy

   Which practice is an example of applying responsible AI in feature engineering?

   • ARemoving all features that are correlated with each other to reduce redundancy
   • BUsing automated feature engineering tools to speed up the process
   • CAuditing engineered features for potential bias against protected groups
   • DIncluding as many features as possible to maximize information
   Reveal Solution Discussion Select an answer first
2. 27expert · hard

   A data science team is building a model to detect fraudulent transactions. The dataset contains transaction amounts, merchant IDs, and customer account numbers. The team wants to engineer features that capture customer spending patterns. What is the most appropriate way to use the account numbers in feature engineering while protecting customer privacy?

   • ARemove the account numbers and only use transaction amounts and merchant IDs.
   • BEncrypt the account numbers and use the encrypted values as features.
   • CUse the raw account numbers directly in the model as a categorical feature.
   • DReplace account numbers with a hash-based token to create a customer ID for aggregating transactions.
   Reveal Solution Discussion Select an answer first
4. 28application · medium

   A data science team is building a model for a healthcare provider. They use multiple data sources, including electronic health records (EHR) and patient surveys. During a review, a stakeholder asks how the 'patient satisfaction' feature was created. The team needs to provide a clear answer. What should they have in place to answer this question effectively?

   • AA verbal explanation from the data scientist who built the model.
   • BA copy of the final training dataset stored in a shared drive.
   • CA data lineage record that traces the feature from its raw source through each transformation step.
   • DA summary of the model's overall performance metrics.
   Reveal Solution Discussion Select an answer first
5. 29expert · hard

   An insurance company is developing a model to set premiums. During data exploration, the team finds that the 'credit score' feature has a strong correlation with the 'claims history' feature. They also notice that credit score is correlated with race. The team must decide how to handle this. Which approach is most appropriate to address the ethical and business risk?

   • ARemove the credit score feature because it is correlated with race.
   • BInvestigate whether the correlation between credit score and claims history is causal or spurious, and evaluate the model's fairness metrics before deciding.
   • CKeep both features because they are both strong predictors of risk.
   • DCombine credit score and claims history into a single composite feature to reduce the number of features.
   Reveal Solution Discussion Select an answer first
6. 30foundation · easy

   Which technique is commonly used to protect sensitive data during feature engineering by replacing original values with pseudonyms?

   • ATokenization
   • BOne-hot encoding
   • CAggregation
   • DNormalization
   Reveal Solution Discussion Select an answer first