Machine Learning and Fraud Prevention


The current generation has the strange privilege of living the greatest workplace change in human history: the industrial revolution based on machine learning. This is a computer’s ability to learn and perform very complex tasks through algorithms and an enormous available processing capacity.

This exponential advancement is due to computers not being programmed to perform tasks, but instead to learn to do them, generally through the use of historical information. This is the paradigm shift.

The adoption of this technology has achieved great improvements in different areas: imaging for diagnosing disease or detecting cardiac conditions; self-driving automobiles and heavy transport; law firms where algorithms learn based on historical information composed by laws, jurisprudence and sentences, along with the results of expertise involved in designing the best defense strategies; retail where algorithms present consumers with products of their interest based on purchases made by individuals of like tastes; leisure activities and social media to capture and individuals attention most of the time; among many others.

Machine learning technologies are also applicable to fraud prevention by providing a great economical value; and analysts that face the challenge of becoming great system administrators and have a hybrid team of humans and machines. Incorporating machine learning is like having Leonel Messi in the fraud prevention team.

These algorithms can learn to differentiate the little details and characteristics that make up a fraudulent transaction. They use historical information from millions of transactions and thousands of variables to achieve the best results in detection levels and false positive ratios.

On the other hand, this world of enormous advances has begun a race between financial institutions over who will lead the era of digitalization. Open institutions, digital onboarding and omnichannelness have opened a door to new opportunities for fraudsters who nefariously exploit technological advancements.

Monitor Plus® uses deep learning, XGBoost and LightGBM algorithms to satisfy both security and customer experience demands. This solution provides coverage to all current needs, is adaptable to the channels and products of the future and incorporates rules-based technology to consider new fraud typologies not foreseen during training. Furthermore, it makes use of communications, making the customer participant in risk management through interactive messaging.


20 vistas