CLS Assembles Data Science Team to Decode FX Volatility

The information services division at CLS Group has assembled a data science team that is currently looking at decoding volatility in CLS data. 

Masami Johnstone, the firm’s head of information services, tells WatersTechnology that she put together the three-member team after she noticed an increased need for information on FX market dynamics from investors. Johnstone joined CLS last year from Euronext, where she was head of buy-side sales and development, and one of her first strategic moves in her new role was to tackle this demand.

Johnstone’s team is currently working on an idea for a product that would provide insight into FX market volatility and liquidity. “Sophisticated investors already use our volume data to predict the direction of FX prices. Our plan is to develop new data products to help our clients better understand FX market dynamics,” she says.

According to its website, CLS processes 500,000 transactions daily worth $1.55 trillion. The daily traded volumes observed by CLS have increased considerably since the spread of the coronavirus, Johnstone says. During the last week of February, volumes hit $2.3 trillion. These figures have continued into March with average daily traded volumes up almost 27% compared to February as a whole. This leaves the team with even more data to digest across a range of three datasets. “The byproduct of this [increased volatility] was a rise of 55% in spot, 15% in FX swaps and 36% in forwards,” she says. 

Each of the data scientists on the team is a physicist or scientist. All were previously part of the broader CLS data analytics team, but were recruited by Johnstone for this specific project. They will now focus on data mining to derive insights from CLS’s datasets and look for relationships between them.

The team is using various machine learning techniques, including deep learning, to find a signal in the massive amount of data they are collecting, and then hone in on specific sets of data and niche areas that can be tested.

Additionally, CLS uses a machine learning automation tool from start-up CausaLens. The software combines multiple CLS datasets and sends out a signal when it discovers a strong correlation between them. CLS currently has five live products that harvest data. Those datasets are then fed into CausaLens’s tool to find correlations.

“Finding a signal that data scientists can focus on is the first step of the machine learning journey,” Johnstone says.

The data scientists can then delve into that specific signal and cut out a large portion of the time they would have spent running the entire process without the help of the CausaLens software. The process is therefore a combination of human decision-making and machine learning, Johnstone says.

“In my mind, this is the most powerful approach because it eliminates the risk of relying only on human beings to determine potential areas of focus,” she says.

Johnstone says that despite the pull of Silicon Valley technology companies, she still sees a good number of young data scientists who are keen to enter financial services. One of the attractions for these young, analytical workers is access to this kind of project, which involves solving complex problems and provides access to unique datasets such as CLS’s, she says.