AWS announces model variable importance for Amazon Fraud Detector

With the new variable importance functionality, Amazon Fraud Detector provides customers a ranked list of model inputs (variables/model features) based on their relative importance to the model’s performance.

By leveraging this information, businesses can better understand their ML models. This information assists in iteratively improve model performance and make more informed decisions, according to the official press release.

First announced in preview at re:Invent 2019 and becoming generally available in 2020, Amazon Fraud Detector is a fully managed service powered by ML that automatically trains, tests, and deploys custom fraud detection ML models based on a customer’s historical fraud data, helping identify potentially fraudulent online activities in milliseconds, the announcement concludes.