DP-100 Question #288: Real Exam Question with Answer & Explanation

Question

Drag and Drop Question You have several machine learning models registered in an Azure Machine Learning workspace. You must use the Fairlearn dashboard to assess fairness in a selected model. Which three actions should you perform in sequence? To answer, move the appropriate actions from the list of actions to the answer area and arrange them in the correct order. Answer:

Explanation

To assess fairness using the Fairlearn dashboard, you must first specify a binary classification or regression model, then select the model feature to be evaluated for bias, and finally choose a metric to measure for disparities across feature groups.

Approach. The correct interaction is to drag the following three actions into the 'Answer Area' in the specified order:

1. Select a binary classification or regression model. Fairlearn's dashboard, designed for fairness assessment, primarily focuses on predictive models, specifically binary classification and regression tasks, where outcomes can be analyzed for disparities across groups.
2. Select a model feature to be evaluated. This step involves identifying the 'sensitive attribute' or 'protected feature' (e.g., age, gender, ethnicity) against which the model's fairness will be assessed. The dashboard will compare model performance for different groups defined by this feature.
3. Select a metric to be measured. After defining the model type and the sensitive feature, you must choose a performance metric (e.g., accuracy, true positive rate, false negative rate for classification; mean absolute error for regression) to quantify and compare potential biases or disparities in model performance across the identified groups.

Common mistakes.

• common_mistake. Common mistakes include selecting incorrect model types or arranging the correct actions in the wrong order. 'Select a multiclass classification model.' and 'Select a clustering model.' are incorrect because the Fairlearn dashboard in Azure Machine Learning is specifically built for evaluating fairness in binary classification and regression models, not directly for multiclass or clustering tasks. Selecting these options would lead to an incorrect answer as Fairlearn's assessment methodologies don't typically apply to them in the same direct manner. Placing the steps in an incorrect sequence, such as selecting a metric before defining the model type or the feature, is also wrong because the evaluation workflow is logical and sequential: first, define the scope (model type), then what to analyze (feature), and finally how to measure (metric).

Concept tested. Fairness in Artificial Intelligence (AI) and Machine Learning (ML), specifically understanding the workflow and capabilities of the Fairlearn dashboard within Azure Machine Learning for assessing model fairness, including identifying appropriate model types, sensitive features, and evaluation metrics.

No community discussion yet for this question.