Rapid and accurate detection of prohibited market practices has become of paramount importance to all business in the financial industry.
Capital market trading has evolved significantly over the past decade, moving on from the open outcry auctions to high-speed trading algorithms placing orders through electronic gateways. Today, most of the market activity across the globe is performed through computer programs which are armed with powerful quantitative techniques such as big data analysis, signal processing and artificial intelligence (AI) technologies.
These tools give their users a significant advantage over ordinary market participants, and has also spawned new and sophisticated forms of market abuse. This trend has created an explosion of data throughout the trading ecosystem, which has resulted in significant compliance challenges. Regulators, on their part, have responded with legislation in an effort to reduce system risk to the financial system and protect its market participants. However, new laws intended to rein in market abuse have presented their own challenges.
One of the biggest challenges that both regulators and compliance staff face is ensuring that the market participants they supervise stay fully compliant with current market regulations. Regulators want to ensure that markets are fair and efficient, and compliance staff understand that the risks of not being able to detect disruptive or manipulative trading behavior can be extremely costly.
As a result, rapid and accurate detection of prohibited market practices has become of paramount importance to all business in the financial industry.
This is where a Chicago, Illinois-based Neurensic comes up with an effective solution. Neurensic's SCORE platform is a detection tool that uses machine learning to identify market abuse and disruptive trading activity like spoofing, momentum ignition, and wash trades.
Speaking to asmag.com, Jay Biondo, Director of Surveillance Products Design at the company said that their detection system is trained on data from actual regulatory cases and investigations in order to accurately identify activity that’s likely to draw negative attention from regulators.
“Our team of data scientists and regulatory professionals work closely together to build detection models that utilize supervised learning techniques on example training patterns while leveraging years of algorithmic trading knowledge and surveillance expertise,” Biondo said. “This adaptive methodology allows Neurensic to continue to improve its capabilities as it becomes exposed to new data over time.”
The technology behind the solution
The technology that powers Neurensic's computational engine is built on top of the machine learning architecture, the H2O framework. H2O is scalable over server clusters and can process trillions of rows of data or more in seconds, given sufficient computing hardware. This allows Neurensic's software to perform surveillance on massive datasets. The technology ports across operating systems, is fully cloud-capable, and is both web browser and mobile app friendly.
What makes it special?
Biondo explained that typical software in the marketplace for trade surveillance utilizes simple rules-based logic with configured parameters, i.e. “send an alert if a trader places an order of size X followed by a cancellation of size Y.” This approach is expensive to implement, time-consuming to review and could be fraught with inaccuracies and false positives.
“Neurensic's approach to trade surveillance and regulatory risk monitoring is totally unique in the industry,” Biondo continued. “Our methodology has the benefit of significantly improving the efficiency of compliance teams by focusing on activity that is the most severe or high-risk, while allowing monitoring of compliance risk trends of the entire organization through interactive visualizations. Since Neurensic’s models are trained via real regulatory data from actual cases and investigations their detection capabilities are able to evolve over time, adapting to changing market dynamics and regulatory mandates.”
And as developments in the field of AI, deep learning and machine learning continue, they will have a significant impact on Neurensic’s solution. The company currently uses only supervised learning techniques when training its machine learning models. However as AI technology continues to develop, Neurensic will potentially be able to use deep learning techniques to develop self-adaptive models, which will increase the precision of its models further and provide users with a better ability to “see around corners” and minimize regulatory risk.