(Changes references to BNY Mellon throughout, from BNY)
Feb 4 (Reuters) - The Bank of New York Mellon Corp said on Thursday it has partnered with Google Cloud on technology that uses artificial intelligence and machine learning to predict when BNY Mellon’s clients’ U.S. Treasury transactions will fail to settle.
Settlement failures occur when two parties buying and selling bonds fail to exchange the cash and securities on time. A large increase in failures can signal market liquidity problems.
The $22 trillion U.S. Treasury market is the largest and most liquid market in the world. Around 2% of transactions fail to settle each day, which can create liquidity shortfalls for clients, BNY Mellon said in a release.
BNY Mellon said its new technology predicts the likelihood that clients, such as banks, will face settlement failure on a trade. The model can predict around 40% of settlement failures for trades that are eligible to be settled on the Federal Reserve’s FedWire Securities Service, with an accuracy of 90%, it said.
The technology analyzes supply and demand and the velocity of trading in specific securities, and it models specific counterparty activity, among other factors, BNY Mellon said. (Reporting By Karen Brettell; Editing by Cynthia Osterman)