How Amazon Fraud Detector works - Amazon Fraud Detector

How Amazon Fraud Detector works

To generate fraud predictions, Amazon Fraud Detector uses machine learning models that are trained with your historical fraud data. Each model is trained using a model type, which is a specialized recipe to build a fraud detection model for a specific fraud use case. Deployed models are imported to detectors, where you can configure decision logic (for example, rules) to interpret the model’s score and assign outcomes such as pass or send transaction to a human investigator for review.

You can use the AWS Console to create and manage models and detector versions. Alternatively, you can use the AWS Command Line Interface (AWS CLI) or one of the Amazon Fraud Detector SDKs.

Amazon Fraud Detector components include events, entities, labels, models, rules, variables, outcomes, and detectors. Using these components, you can build an evaluation that contains your fraud detection logic.

For additional information about Amazon Fraud Detector concepts and how they operate, see Amazon Fraud Detector concepts.

Amazon Fraud Detector workflow

The steps for creating a model, building a detector, and getting fraud predictions include:

  1. Define the event you want to evaluate for fraud.

  2. Gather historical event data (training data).

  3. Create a model version (trained model) using a model type.

  4. Create a detector that includes the model version and decision rules.

  5. Send events to Amazon Fraud Detector and get a fraud prediction.