Posted On: Jul 9, 2021

We are excited to announce that Amazon Fraud Detector now includes model variable importance values with every new fraud detection machine learning (ML) model to provide customers with more insight into their model’s performance.

Fraud Detector is a fully managed service that makes it easy to identify potentially fraudulent online activities, such as the creation of fake accounts or online payment fraud. Using ML under the hood and based on over 20 years of fraud detection expertise from Amazon, Fraud Detector automatically identifies potentially fraudulent activity in milliseconds—with no ML expertise required. Fraud Detector customizes every ML model based on each customer’s unique business.

With model variable importance, Fraud Detector now provides a ranked list of model inputs based on their relative importance to the model’s performance. This information helps customers better understand their ML models and makes it easier to iteratively improve model performance. For example, if customers observe that their model’s performance is driven by IP address, they may choose to include other IP address-related attributes in their model. Model variable importance values are included with every new Fraud Detector model at no additional cost, and can be accessed via the AWS console, AWS CLI and AWS SDKs.

The model variable importance feature is available today in all AWS regions where Fraud Detector is available: US East (N. Virginia), US East (Ohio), US West (Oregon), Europe (Ireland), Asia Pacific (Singapore) and Asia Pacific (Sydney). For more details, see the Fraud Detector user guide.