Proof of concept: machine learning in the fight against bank fraud
Mexico is the country with the most bank fraud in the world
Machine learning (ML) has become commonplace in our day-to-day life. The clearest example is when we open an app or a search engine and something we had been looking for appears, as if by magic. This is no coincidence. What lies behind this type of experience is an innovation driven by analytics, in which we all participate, creating an impressive source of information through the use of digital everyday platforms (websites, social media, etc.). This technology, which belongs to the field of artificial intelligence (AI), is nothing more than a system that can learn patterns, trends and relationships in an automated manner.
Thanks to the analytical power of machine learning, we can follow the digital footprints left behind by users in their activities and define parameters to detect actions that were harder to predict in the past. This tool allows companies to design scenarios, highlight anomalies, identify high-risk customer profiles and even prevent transactions that could mask fraud.
Mexico, the global leader of bank fraud
According to a report by a Mexican newspaper and a study by MasterCard, Mexico is the country with the most bank fraud in the world, surpassing European nations by as much as five times. The results reveal that this type of crime is closely linked to identity theft and to global embezzlement organizations.
In December 2018 alone, credit card fraud rose by 69%, which poses a risk to users and is an unprofitable business for companies.
The most common types of fraud are unauthorized ATM transactions (unidentified activity) at 33%, and card cloning at 35%. Most fraud is reported through e-commerce, with more than 4 million cases each year and $250 million embezzled. This type of cyberfraud involving stolen cards or information taken from customers rose by 25%.
Although fraud is usually done using skimmers (devices that copy data from the magnetic bands on cards), other common methods include creating random passwords, using malware and hacking online shops.
Machine learning to detect fraud
Nae is performing proof of concept (PoC) tests with a number of banks and financial institutions by applying machine learning through DataRobot, one of the top solutions in the world. The process consists of searching for, among millions of algorithm combinations, the ideal machine-learning model for predicting fraudulent transactions and suspicious activity.
These proof of concept tests also introduce customers to predictive analysis tools. Thanks to the intuitive, web-based interface, anyone can interact with the platform.
Regardless of their machine learning experience, users can select models and parameters to be assessed by the system, and then implement solutions that support decision-making.
Nae is a certified DataRobot partner for artificial intelligence and machine learning solutions aimed at businesses. The company specializes in automating the data science workflow in order to recommend algorithms and build predictive models.
Casos de estudio