LONDON (Reuters) – The use of machine-learning has picked up across the financial services industry, although problems such as data quality continue to dog its progress, a study by Refinitiv has found.
FILE PHOTO: A chart is displayed behind a trader on the floor of the New York Stock Exchange (NYSE) shortly after the opening bell in New York, U.S., March 26, 2019. REUTERS/Lucas Jackson/File Photo
In a report, Refinitiv, the financial data provider, said more than 90 percent of the organizations it surveyed had either deployed machine-learning in multiple areas of the organization or have made a start in some pockets.
Machine-learning refers to the use of algorithms and statistical models in financial markets without using human directions and instead relies on patterns to make choices.
While the initial driver of such technologies was the automation of repetitive tasks, the survey found that the top applications were in the areas of risk avoidance, generating trading and investment ideas and analyzing performance.
Conducted via 447 telephone interviews of senior executives and data-science practitioners across various financial services firms, the survey also found the quality of data as the primary barrier to machine-learning adoption.
Machine-learning has long been the mainstay of deep-pocketed hedge funds, which have combined complex algorithmic strategies with financial data to make big bets on markets.
But with the growing use of cloud computing and the constant pressure on banks to reduce costs, machine-learning techniques have seen a greater acceptance among banks.
“Thanks to parallel computing and cloud computing, we are seeing the playing field being slowly leveled in terms of machine-learning strategies,” said Tim Baker, global head of applied innovation at Refinitiv.
The survey also found foreign exchange ranked a distant fourth in terms of structured data by asset class with stocks, fixed income and derivatives the top three.
Foreign exchange markets with a myriad trading platforms and opaque over-the-counter trading format was a particular problem area for data scientists when it came to applying machine-learning strategies.
“Standing in the way of machine-learning adoption in the FX market is a potent combination of data science, engineering and technology management problems,” said Matthew Hodgson, CEO and founder of Mosaic Smart Data, a data analytics company.
Reporting by Saikat Chatterjee; Editing by David Evans