Posted On: Mar 30, 2021
Amazon SageMaker Autopilot, which makes it easy to create highly accurate machine learning models, now provides a model explainability report generated by Amazon SageMaker Clarify, making it easier to understand and explain how the models you create with SageMaker Autopilot make predictions. Explainability reports include feature importance values so you can understand how each attribute in your training data contributes to the predicted result as a percentage. The higher the percentage, the more strongly that feature impacts your model’s predictions. You can download the explainability report as a human readable file, view model properties including feature importance in Amazon SageMaker Studio, or access feature importance using the SageMaker Autopilot APIs.
By understanding how your model makes predictions, you can make more informed business decisions. For example, you can verify that your model is behaving as expected by confirming that attributes with a high importance value represent a valid signal for predictions in your business problem. With model explainability reports, you can remove attributes that are less important to create models that make predictions faster. You can check the fairness and accuracy of your model by identifying the attributes from which you want to remove bias and confirming if they have low feature importance.
Explainability reports are now available in SageMaker Autopilot in all regions where SageMaker Autopilot is available. To get started, visit our documentation and webpage.