Shallow or deep? Detecting anomalous flows in the Canadian Automated Clearing Settlement System using an autoencoder.

Published: February 25, 2020

Payments Canada has released its latest discussion paper, Shallow or deep? Detecting anomalous flows in the Canadian Automated Clearing Settlement System using an autoencoder. This paper shows how unusual payment flows in the Automated Clearing Settlement System (ACSS) can be detected by an artificial intelligence (AI) application. AI is the theory and development of computer systems able to perform tasks normally requiring human intelligence. In order to comply with the prominent payment system (PPS, which are based on the Principles for financial market infrastructures, PFMIs) guidelines, the ACSS operator needs to manage its risks. Participants of ACSS have daily payment flows among each other stemming from different retail payment instruments. In case a participant changes its behaviour in this system or faces liquidity or operational problems, it can have a severe impact on the other participants in the system. Having a tool available that is able to detect potential changes in behaviour or potential liquidity problems at the participant level automatically, would help the operator to get a better understanding of what is happening in its payment system. The sooner potential problems are known, the more time there is to take adequate measures.