MLA-C01 Question 228
Select 2You are a Machine Learning Engineer using Amazon SageMaker Clarify to evaluate the fairness and bias of your model and training data. Which metrics provided by SageMaker Clarify can help you analyze bias and disparity in your dataset and model predictions?
- A
Difference in Positive Proportions (DPP)
- B
Disparate Impact (DI)
- C
Shapley Value-based Feature Importance
- D
Precision and Recall Rates
- E
Class Imbalance Percentage
Show answer and explanation
Correct answers: A, B
Explanation
SageMaker Clarify provides metrics such as Difference in Positive Proportions (DPP) and Disparate Impact (DI) to help identify and quantify bias in datasets and models. These metrics are crucial for evaluating fairness across different demographic or protected groups. Other options like Shapley values and general evaluation metrics (e.g., Precision and Recall) are unrelated to bias-specific insights.
- A. Correct.
Difference in Positive Proportions (DPP) is a metric provided by SageMaker Clarify to measure fairness by comparing the ratio of positive outcomes across different groups.
- B. Correct.
Disparate Impact (DI) is another fairness metric provided by SageMaker Clarify that evaluates the ratio of favorable outcomes for a protected group to that of an unprotected group.
- C. Incorrect.
Shapley Value-based Feature Importance is used for explainability in SageMaker Clarify, not for bias or fairness measurement.
- D. Incorrect.
Precision and Recall Rates are general performance evaluation metrics for ML models but are not specifically used in SageMaker Clarify for bias analysis.
- E. Incorrect.
Class Imbalance Percentage is a dataset property but is not a SageMaker Clarify metric for bias or fairness evaluation.