The number of rule changes global financial institutions must adhere to every day has trebled since 2001 to an average of almost 200 revisions a day. As a result, the amount of data fed back to regulators has multiplied to unprecedented volumes.
Using a concoction of artificial intelligence (AI) algorithms, MindBridge has developed an AI platform that automates the ingestion of datasets, analysing them to produce an individual ‘risk score’. The platform then presents the results via a user interface, which regulators can then use to search for non-compliant activity.
MindBridge, winner of Central Banking’s FinTech & RegTech Global Award for Best Machine Learning Solution For Regulatory Compliance, was established in Canada in 2015 to find a way of using different AI tools – notably machine-learning algorithms – to detect anomalies in large datasets.
The MindBridge user interface allows central banks to drill down into individual transactions, both in isolation and in comparison to peer firms, enabling the user to explore a time series of each variable in correlation to the industry average. A patent-pending Natural Language Processing search engine provides regulators with search capabilities to quickly and accurately identify the information they want.What is different about MindBridge’s AI platform is its lack of reliance on rules‑based analytics, which it says restricts the amount of information regulators can get from datasets. Instead, the platform uses a combination of business rules, statistical models, machine-learning analytics and other AI‑based algorithms to detect anomalies in data.
The promise of MindBridge’s technology is evident in the clients the firm has acquired over the past 12 months. Last year, the company was involved in the Bank of England’s (BoE’s) Fintech Accelerator. Over eight weeks, MindBridge built a bespoke tool for BoE’s Data and Statistics Division, which collaborated with MindBridge to learn how the software analysed the data and detected anomalies. It was then put into practice to identify anomalies in anonymised credit union datasets.
Subsequently, BoE looked at the versatility of the platform to provide data visualisation and preparation techniques for larger numeric and transaction‑level datasets – particularly to explore the potential of machine learning to assist the way it conducts plausibility and validation checks on different types of datasets.
Once the project was finished, BoE signed a contract with MindBridge allowing the use its technology to analyse its high‑frequency transaction data, which is being used to help calculate the BoE’s reformed sterling overnight index average (Sonia) rate.
Payments Canada is using the MindBridge platform to analyse large-value transfer system transaction and credit limit data. This initiative further showcases the broad capability of the MindBridge Ai platform by quickly identifying anomalies and deriving valuable insights within narrow datasets.