Exam AWS Certified Machine Learning - Specialty topic 1 question 226 discussion

As the model has been already trained and deployed, I will go with C. because B (SageMaker Debugger) is used at training time

C is right B and C are possible solutions, but the question requested the MOST operationally efficient manner

SageMaker Clarify provides tools to help ML modelers and developers understand model characteristics as a whole prior to deployment and to debug predictions provided by the model after it’s deployed Option B is not recommended because retraining the model with SageMaker Debugger and configuring Debugger to calculate and collect Shapley values is time-consuming and may not be operationally efficient

It seems that both B and C are possible answers. SHAP baselines can be provided by both. The scenario says nothing about concern of bias, so perhaps Clarify is overkill? This post from AWS seem to be specifically addressing this case, and uses SageMaker Debugger. https://aws.amazon.com/blogs/machine-learning/ml-explainability-with-amazon-sagemaker-debugger/

Selected Answer: B Retrain the model by using SageMaker Debugger. Configure Debugger to calculate and collect Shapley values. Create a chart that shows features and SHapley Additive explanations (SHAP) values to explain how the features affect the model outcomes. While A, C, and D are all options for explaining the model's behavior, the most efficient way to meet the bank's requirements is to use SageMaker Debugger to calculate and collect S Shapley values for each prediction. This allows the data science team to easily explain why the model denied the promotion to certain customers. SageMaker Debugger also provides built-in integration with SageMaker Studio, which enables data scientists to visualize the Shapley values and other debugging information through a user-friendly interface.

I think it's D. Model monitor automatically integrated with Clarify

B. Retrain the model by using SageMaker Debugger. Configure Debugger to calculate and collect Shapley values. Create a chart that shows features and Shapley Additive explanations (SHAP) values to explain how the features affect the model outcomes would be the most operationally efficient way to meet the requirement of explaining why the model denies the promotion to some customers.

Clarify is the best solution. The key is training data https://www.amazonaws.cn/en/sagemaker/clarify/

Its between B and C SageMaker Clarify is used to promote transparency and accountability in machine learning models. Thats what we are looking for why model denies promotion to some customers

"Explain individual model predictions Customers and internal stakeholders both want transparency into how models make their predictions. SageMaker Clarify integrates with SageMaker Experiments to show you the importance of each model input for a specific prediction. Results can be made available to customer-facing employees so that they have an understanding of the model’s behavior when making decisions based on model predictions." https://www.amazonaws.cn/en/sagemaker/clarify/