Header newsletter test

[Experts Opinion] Using Machine Learning To Fight Against Fraud

In: Fraud

quote-iconToday we are going to delve further into this topic with two of our Data Science Project Managers. They will be explaining another side of HiPay Sentinel: using Artificial Intelligence in fraud management with our tool Smart Decision.quote-icon


  • Can you tell us a little bit more about the development of Artificial Intelligence at HiPay?

Take a look at the e-commerce today. According to a FEVAD report last year, online sales increased more than 14% in the first trimester of 2017. This kind of growth is frequently accompanied by a higher fraud risk that merchants have to deal with.

Usually, many payment providers combat fraud by estimating transaction risk level based on rules that then result in those transactions being given a score. As Virginie explained, this is useful as a first filter for transactions. Whether authorized or blocked, these transactions require further analysis by the e-merchant. This means performing manual reviews that require both skills and resources that the merchant doesn’t necessarily have. However, this is a long process that erodes the customer user experience and puts customers’ payments on hold until processing.

This is why we started working with Artificial Intelligence. The objective of this technology is to make machines ‘intelligent’ in order to quickly solve problems. Machine learning is a technique that accomplishes this. With Machine Learning at the heart of our new product HiPay Sentinel, we use our technical expertise to help our customers by having transactions reviewed against algorithms that are able to learn and adapt. The idea is to have the score module complement our experts’ knowledge and Smart Decision.


  • Can you give us more details on Smart Decision, our Machine Learning module integrated into HiPay Sentinel for processing ‘challenged’ transactions?

As you’ve seen, ‘challenged’ transactions need to be reviewed after being processed by our anti-fraud module. The first step using Machine Learning is to ‘teach’ the algorithms by analyzing the data history of each of our merchants. In general, we want to define criteria based on fraud patterns. Then, with the amount of data at our disposal, we are able to successfully determine how much risk a transaction poses. This allows statistical models to be created. If we evaluate the relevance of our modules from a statistical point of view, we are able to take the merchant’s business dimension into account. The aim is to maximize the number of true positives (fraudulent transactions blocked by the system) while minimizing the number of false positives (non-fraudulent transactions that are blocked).

To summarize, Smart Decision is a permanently auto-learning technology for automatic processes. Therefore, Machine Learning allows us to evaluate the level of risk for transactions that need reviewing to make the best decision in real-time.

  • Do you know if you’ll stop here in the development of this technology? What is your vision for the coming months?

Our first steps in this project have increased our comfort level using Artificial Intelligence and in particular Machine Learning that allows us to improve our product quality. We will continue pursuing this path with HiPay Sentinel… but it doesn’t stop there! We are currently working on several projects that will allow us to help merchants grow their business activity by implementing new and innovative services adapted to their needs.


HiPay’s activity doesn’t stop with machine learning. Follow our blog to keep posted on our latest happenings!

Tags: Fraud

    Link Articles