Posted On: Feb 6, 2023
Today, Amazon Fraud Detector (AFD) announced the launch of the Cold Start feature. Now customers can start training a sign-up or a transaction frauds detection model with minimal historical-data. Up to now, AFD customers were required to provide 10K+ labeled events with at least 400 examples of fraud to train a model. With the release of Cold Start only 50 labeled fraud events and 50 unlabeled events are necessary. The new feature introduces intelligent methods for treating your unlabeled data and optimizes model training with small datasets.
The most significant obstacle for any organization looking to leverage machine learning in its business is the requirement to have rich, consistently formatted historical data. Lack of significant historical data may be regarded as a data ‘cold-start’ scenario for training a ML model. Now with AFD, customers can get started quickly with a quality fraud detection model when there is only limited model training data available. Customers can then start iterating on fraud tagging and continuously re-training their model with growing datasets to increase model performance.
Amazon Fraud Detector (AFD) 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 AFD automatically identifies potentially fraudulent activity in milliseconds.
This feature is available in all regions where Amazon Fraud Detector is available: US East (Ohio), US East (N. Virginia), US West (Oregon), Europe (Ireland), Asia Pacific (Singapore) and Asia Pacific (Sydney). To learn more about Amazon Fraud Detector, visit our product page.