MLA-C01 Question 225
Select 2You are using Amazon SageMaker Clarify to analyze the fairness and explainability of your machine learning model. Which metrics provided by SageMaker Clarify can help you assess bias in the training data and the model?
- A
Class Imbalance Ratio
- B
Disparate Impact
- C
Shapley Values
- D
Demographic Parity Difference
- E
Feature Importance Scores
Show answer and explanation
Correct answers: B, D
Explanation
To assess bias in training data and models, SageMaker Clarify provides specific metrics such as Disparate Impact and Demographic Parity Difference. These metrics help measure fairness by comparing outcomes across different groups. While metrics like Shapley Values and Feature Importance Scores are useful for explainability, they do not directly assess bias.
- A. Incorrect.
Class Imbalance Ratio is not a metric provided by SageMaker Clarify directly, but it may be inferred indirectly by analyzing datasets. SageMaker Clarify focuses on metrics specifically designed to measure bias and explainability.
- B. Correct.
Disparate Impact is a bias metric provided by SageMaker Clarify that assesses the ratio of favorable outcomes for different groups, helping measure bias in predictions or training data.
- C. Incorrect.
Shapley Values are used to explain model predictions and are part of explainability tools in SageMaker Clarify, but they do not directly assess bias in the data or model.
- D. Correct.
Demographic Parity Difference is a bias metric provided by SageMaker Clarify that measures the difference in positive outcome rates between groups, making it relevant for bias analysis.
- E. Incorrect.
Feature Importance Scores are not bias metrics but are used to understand the influence of features on predictions, which is part of explainability rather than bias assessment.