Machine learning and artificial intelligence make it possible to reduce the risk when granting credits, detecting signs of non-payment and combating the challenges posed by cybersecurity.

The application of advanced technology is increasing efficiency in all sectors of activity. And in this process of digital transformation, financial entities may be the most dynamic in the adoption of all kinds of advances.

Machine learning is one of the technologies that can contribute most to a better development of banking business. “Machine learning techniques provide greater predictive power in banking credit models,” commented Sergio Lloret, an account manager at the AIS Group consultancy, at the last National Credit Congress in Spain.

He explained that the results obtained in recent projects show that the level of success in credit granting and monitoring models improves between 25% and 50% when using machine learning algorithms instead of traditional techniques. Lloret said that these results are reflected in the so-called Gini Index, “which goes from 50% or 60% to be in ranges above 90%, which is a percentage of high success.”

In this way, the joint application of algorithms and techniques of artificial intelligence (AI) and machine learning to risk models allows us to multiply their success. And this is fundamental for the financial sector because it helps entities adjust to their level of risk appetite .

In addition, these techniques allow monitoring of the portfolios and detect the signs of a possible default earlier , and may act before it occurs. Ramón Trías, president of AIS Group, explained that this ability to anticipate “is especially useful in the new framework imposed by IFRS 9 (NIIC 9), as a change in the state of operations can impact negatively in the income statement, by causing the obligation to increase the level of reserves. “

On the other hand, it influenced the role of AI in cybersecurity. “Each time the malware is more advanced and even uses AI techniques. It is no longer limited to knocking down the systems of the companies it affects. Now there is another type of more elaborate and organized attacks: progressive information thefts, massive and automated attacks with bots, information encryptions, user identity theft, etc. It is necessary to resort to very advanced systems to detect and stop these smart malware practices before it is too late, “he said.

Also, Trías stressed that we live a moment of “explosion” of these technologies , which has been working for decades. However, now several factors that are driving its development are combined. First, he highlighted the advances in hardware and software, which have allowed to multiply the speed of computing. “A computation that used to take 4 months, now is ready in just 20 seconds, which is an industrial time,” he said. And this increase in capacity, combined with the emergence of big data, allows us to work with more variables, improving the power of prediction.“Having more information does not mean using it all. You have to know which variables are the most relevant and really add value to the algorithm. The AI ​​is a great help, but the role of the expert is still key to direct the work and constant learning of these advanced systems, “he said.