An IS auditor is reviewing a data set used by a university to train a predictive machine learning (ML) model. Which of the following MOST likely indicates risk that the model could not process all data and make necessary correlations?
Answer(s): C
A numeric field such as Final Grade Percent stored in object (string) format prevents the model from interpreting it as quantitative data, limiting its ability to process the field correctly or make meaningful correlations during training.
An IS auditor for a veterinary clinic was informed that the dog breed categorical variable within the data set is necessary for the predictive machine learning (ML) model conclusions. Which of the following introduces the MOST risk in this situation?
Ordinal label encoding assigns numerical values that imply an order or ranking among dog breeds, which is incorrect and introduces significant modeling risk. This false ordering can distort relationships learned by the ML model and lead to invalid or biased predictions.
Which metric should an IS auditor review to evaluate issues with data collection that could impact the AI model training?
Answer(s): B
A high percentage of missing values indicates problems in data collection and can significantly impair model training by reducing data quality, completeness, and reliability.
Which of the following is the BEST way to ensure data fed into an AI model aligns with business objectives?
Documenting data input requirements ensures that only data fitting the business objectives, definitions, and intended use cases is fed into the model. This alignment prevents incorrect assumptions, irrelevant inputs, or misuse of data that could undermine model effectiveness.
A newly deployed fraud detection model is misclassifying transactions due to inconsistent formatting in the data stream. Which of the following is the IS auditor's BEST recommendation to ensure consistent model performance?
Answer(s): A
Defining and documenting technical specifications for incoming data ensures that the data stream is formatted consistently before reaching the model. This directly resolves inconsistencies that cause misclassification and stabilizes model performance.
Post your Comments and Discuss ISACA AAIA exam prep with other Community members:
Our website is free, but we have to fight against AI bots and content theft. We're sorry for the inconvenience caused by these security measures. You can access the rest of the AAIA content, but please register or login to continue.