BIS prototypes ‘game-changing’ bank risk prediction platform

BIS news platform

The Bank of International Settlements (BIS) has announced it has successfully prototyped a ‘game-changing’ news media analysis platform designed to forecast risks to individual banks and gauge wider economic sentiment.

Dubbed ‘Ellipse’, the prototype platform extracts and then analyses, in real-time, unstructured data from news events and market insights (including management reports, financial and regulatory disclosers and external credit assessments).

Using machine learning, this data is then combined with structured regulatory data and analytics (from regular reporting by banks provided to corporate regulators) to enable watchdogs to prompt banks with early warning of potential risks to their operations as well as risks to the banking system more widely.

Co-developed with the Monetary Authority of Singapore (MAS), a key priority for the developers was to prove that these combined datasets could be integrated into a single platform (a “customisable dashboard user interface”) to assist regulatory authorities in identifying potential risks and ultimately improve their supervisory decisions.

The project was also backed by the Bank of England, the International Swaps and Derivatives Association, Accenture and FNA.

In designing the project, the BIS noted that financial supervisors still rely heavily on regulatory reporting to inform them of potential risks. However, it recognised several challenges with this approach which ultimately impact the quality of their risk assessments.

Firstly, information sourced from regulatory reports is often template-based, making it difficult to share and extract. Secondly, because data is often sourced from legacy systems, it can be inconsistently arranged and described. Perhaps most critically, however, is the timeliness factor, with regulatory reports submitted infrequently (only every month or quarter, for instance) meaning that information is “dated and backward-looking”.

Project Ellipse, the BIS wrote, effectively set out to demonstrate that “real-time insights using advanced analytics could be derived from large volumes of structured and unstructured data… to make risk correlations and analyse sentiment, alerting supervisors of issues that may need further investigation”.

The initial proof of concept was limited in scope, the BIS acknowledged, intended to determine whether data could be “represented in a way that allows programmable code… [to] generate the reporting of regulatory metrics”.

The second phase of the project explored whether the BIS could build a single, one-stop-shop platform to give regulatory authorities “‘on demand’ access to timely and integrated sources of data” that could support and inform their supervisory decision-making.

The platform utilises machine learning and natural language processing on unstructured and granular reporting data to identify risk correlations and sentiment analysis; it then alerts supervisors in real-time to any issues that may need further investigation.

“We…wanted to show that a single platform could be capable of acting as a ‘one-stop-shop’, where supervisors could find regulatory information quickly and could be guided by insights generated from the analytics running on the platform,” the BIS noted.

Acting head of the BIS Innovation Hub, Ross Leckow, recognised the need for regulators to have on hand accurate and timely information to assess emerging risks.

Ellipse, he said, has the potential to be a “game-changer” for regulators, giving them access to “more and better data, structured and unstructured, with greater predictive insights than ever before”.

He added that the system “can be scaled to provide real-time analysis on a national or cross border supervisory basis”.

Hern Shin Ho, deputy managing director (Financial Supervision) at the MAS praised the Project as a clear demonstration “that collection and use of such datasets need not be prohibitive, but can be codified, efficient, cost-effective and potentially scalable even on a cross border basis”.

“MAS is adapting the prototype for our own supervisory needs. I hope other supervisors will similarly find it useful and look forward to further joint initiatives to develop common SupTech solutions for supervisors.”

The Ellipse prototype is the first to be published on BIS Open Tech platform, designed by the BIS to share statistical and financial software as public goods to promote “international cooperation and coordination”.