Amex is confident that its new ML-based fraud detection model will remain a leader in the industry. But, it does have to consider costs. This article discusses the costs of the new system. Amex is also confident that it can stay one step ahead of the competition. What are the other benefits of the new model? What can Amex expect from it? We’ll be posting more details as they become available.
Amex ML-based fraud detection model
Initially, the Amex team had planned for the Gen X build to last about three months, but it ended up taking almost a year because of Covid-19. However, when the new system finally arrived in January of 2021, the company attributed it to the extra time it took to fine-tune the model. In a sense, the new generation model is like a Choose Your Own Adventure novel, with every decision having a path-dependent conclusion.
AmEx has launched both a consumer and business checking account. In October, the company began offering both accounts. Despite this early success, the new model is designed to identify suspicious patterns and odd behavior. The model will also catch up on a recent trend, which is allowing fraudsters to take over profiles and apply for accounts in a customer’s name.
Another benefit of this new ML-based fraud detection model is that it can recognize fraudulent transactions without the need for human intervention. As with any rule-based system, there is a limit to the number of rules and transaction features that can be used to determine fraud cases. Besides, rule-based systems require a lot of manual development and maintenance.
While it’s not as simple as a purely AI-based system, the Amex ML-based fraud detection model has the advantage of a clearer goal compared to other financial institutions. Amex’s data is much less fragmented than Capital One’s, making it easier to transfer models between companies. That means more accuracy for consumers, and greater efficiency for the company.
Gen Amex Mlbasedfield
Because fraudsters are more likely to target new financial products, the Gen Amex ML-based fraud detection platform is more advanced than its rule-based counterpart. Its algorithms can better detect fraud and improve the overall accuracy of fraud detection. This new model can also help the financial services industry compete with competitors using ML-based solutions.
As the technology behind the new Gen Amex fraud detection model is still evolving, Amex executives are being coy about the technology that will be used in its new system. However, they are preparing to roll out its tenth major global fraud detection model. The Gen Amex model uses generative adversarial networks (GANs) and sequential recurrent neural networks to identify fraudulent transactions.
The new system is also a solution to the current challenge of over-simplifying fraud detection. Its approach involves analyzing past transactions and recording a detailed description in the form of feature vectors. This approach does not attempt to understand fraud directly. Instead, it records past transactions. Then, it generates a logical explanation of those transactions in the form of a feature vector.
To make this approach more efficient and effective, Amex has gathered a team of experts from various backgrounds. The Ealax team is led by Edgar Lopez-Rojas, a PhD in Computer Science who has developed simulation methods for fighting financial fraud. Edgar has previously served as a researcher at NTNU and Simudyne, where he developed financial fraud analytics techniques. He also served as a member of the ISO Standards Norge committee on Fintech SN/K 250 – Bank and Financial Services. Among the speakers were Juan Sebastian Cardenas-Rodriguez, a software developer intern at Ealax.
Amex is confident that its new model will help it to remain a leader in the field of fraud detection
One problem with fraud detection models is that they tend to be overly sensitive, and this can result in a halting of legitimate transactions. Fortunately, AmEx does not always stop a transaction after a fraud alert is triggered. Instead, it may send a message asking the customer if they made the purchase or took any action as a result of the alert. Ultimately, this false positive information and customer feedback feeds back into the fraud model.
READ MORE: Upgrade Your HP EliteBook X360 1040 G6 Processor