Amid widespread hype around machine learning, trade surveillance is one area where practical applications and success are already a reality. As advances continue, don’t miss out.
Machine learning holds huge promise for the evolution of trade surveillance and monitoring, some of which is already being fulfilled. While not all surveillance activities will benefit from machine learning, and some technology remains experimental at best, established machine-learning-powered platforms are now helping banks’ 1st, 2nd and 3rd lines of defence strengthen their risk and control frameworks right now.
Cut through the noise
As banks seek to cut both cost and complexity from their trade surveillance functions, intelligent automation can help with both, believes regtech software provider Eventus Systems, which includes machine learning as a standard offering with its Validus surveillance and market risk software platform.
Machine learning on its own, however, is no panacea to the challenges that the increasingly stretched surveillance functions face. To identify suspicious trading practices or market abuse such as spoofing or layering in the most efficient way possible, banks need to take multiple approaches to alert generation and resolution, casting a wide net from the outset and using a platform that combines machine learning with other forms of automation, argues Scott Schroeder, Eventus’s global head of sales.
Such a holistic approach can help them cut through the noise to identify the most reliable alerts while reducing time and money spent triaging and closing the thousands of spurious false positives that many surveillance practitioners are obliged to manually click and comment on every day. It also means that institutions can probe more deeply where necessary and take resolution action more quickly.
“Integrating machine learning into our platform has created a more efficient process for customers to take millions or sometimes billions of messages and reduce them to a manageable number of actionable alerts,” says Schroeder.
Customised and self-improving
The risk appetite, business functions and manpower availability vary from customer to customer, with some having a higher tolerance for false positives, while others want to investigate only specific types of alerts. For example, a market making firm will generate a high cancel-to-order rate that in a different scenario might be indicative of layering. For these reasons, Validus allows firms to define and adjust their risk profile and parameters, which can then be further honed by machine learning technology.
Eventus also employs a continuous improvement model under which customers – either individually or as a group – can provide scores for alerts, which the machine then learns from to improve the accuracy of its assessments.
Satisfying RTS 6 and algo monitoring
As ballooning algorithmic trading volumes continue to attract the attention of regulators, banks and other financial institutions can also benefit from intelligent automation to satisfy the increasingly stringent regulatory demands.
For example, under the EU’s MiFID II financial industry reform legislation, investment firms that engage in algorithmic trading are required to perform an annual self-assessment and validation process under the RTS 6 standard.
Eventus is therefore anticipating significant growth in demand for tech solutions that can simultaneously support banks’ surveillance capabilities and help them satisfy MiFID II’s RTS 6 algorithmic trading control requirements. “So far, we’re the only provider we’ve seen that can do both,” Schroeder says.
Cryptocurrencies up next
Looking ahead, firms that participate in the massive foreign exchange trading market, which will continue to face more onerous regulatory requirements for surveillance, will need tech solutions to help them do so. Intelligent automation will be crucial in helping them to boost surveillance capabilities. Similarly, regulators such as the Securities & Exchange Commission, Commodity Futures Trading Commission and the Financial Conduct Authority are taking a keen look at cryptocurrency trading and markets.
Broker dealers, proprietary trading groups and even buy side firms that participate in cryptocurrency trading will likely need to enhance their surveillance solutions if, as expected, these regulators eventually mandate that they monitor their crypto activities.
“It’s moving in that direction and we think we’re well-positioned for when that happens,” Schroeder asserts.
Don’t have FOMO over ML
While machine learning and AI are already having a tangible impact on banks’ surveillance capabilities, a degree of patience and perspective is important.
Unsupervised machine learning, for example, will eventually stretch the boundaries of what compliance and monitoring functions are capable of and smaller tech providers are still experimenting in this arena. The fragmented nature of financial markets, with a multitude of different contract types and trading strategies in use, means that supervised learning still offers the best and most reliable practical applications for trade surveillance at scale, Schroeder believes.
The surveillance function is only at the start of its journey with machine learning, but progress is developing quickly. Firms that invest in surveillance systems and embrace intelligent automation now will be ahead of the pack and well-positioned to take advantage of advances in machine learning as they occur. To those that wait, it won’t be a case of FOMO (fear of missing out), it will simply be MO: missing out.
Eventus Systems’ Validus platform provides sophisticated market surveillance and financial risk capabilities, enabling clients to solve some of the most pressing regulatory challenges. Visit eventussystems.com.