Northfield Information Services Launches Risk Systems That Read®
Northfield Information Services’ today announced the launch of Risk Systems That Read® – a service that augments Northfield’s current risk models.
Unlike traditional risk models that rely only on statistical data from the past, models augmented by Risk Systems That Read quantify risk based on analysis of the same news text that investors are receiving now. Investment professionals get a more complete and accurate understanding of risk in financial markets.
Risk Systems that Read will be useful in a number of applications, particularly for those in the hedge fund space, as well as fund managers and asset owners. Algorithmic trading, including “high frequency” and optimal execution algorithms, represents another area of great potential importance. This new approach also offers daily updating of fixed income credit risk and derivative counterparty risk.
“With Risk Systems That Read we are measuring how the present is different from the past and, therefore, how the near future is also likely to be different from the past — something missing from nearly all financial models,” says Northfield President Dan diBartolomeo. “Unlike risk models that rely only on historical data, Risk Systems That Read improves risk forecasts based on the volume and content of financial news, providing users with a more complete and accurate view of risk. It delivers the stability of longer-term model forecasts enhanced by rapid adaptation to changing market conditions, at the individual security level across the full universe of asset classes.”
Risk Systems That Read is a fully complete application of “Big Data” techniques. Financial institutions do not need their own team of data scientists and domain experts to take advantage of the process.
Here’s how it works: Northfield’s partner, Alexandria Technology, a pioneer in machine learning, has developed classification algorithms that mimic the way a research analyst would evaluate news: capturing the who (entities), what (topics or events) and importance of a story (sentiment). On a typical day, Alexandria analyzes more than 5,000 news articles from Dow Jones & Co., including premium publications such as the Wall Street Journal and Barrons.
The classification takes less than thirty-milliseconds, according to Alexandria President, Dan Joldzic.
“Using machine learning, we’re able to analyze text far better than simply looking for good or bad words,” says Joldzic. “Our engine is able to model language, not just the words or phrases, meaning we can capture the true context of information and turn that into an accurate summarization of an article. Northfield recognizes this advantage with us and has done incredible work in applying our output into their risk models.”
From there, the data sets head to Northfield who applies the Risk Systems That Read analytics and updates existing models. Those models are sent to clients who subscribe to Northfield’s new horizons service overnight.
diBartolomeo is convinced that Risk Systems That Read “will be the biggest step forward in risk modeling for asset management since the creation of the multi-factor risk model in the 1970s.”
Kevin Shea, CEO of Disciplined Alpha LLC is also optimistic about the potential for Risk Systems That Read. "Northfield, a firm that is always on the cutting edge of quantitative research, has recently incorporated news analysis into their risk models,” Shea says. “We are pleased to see it has collaborated with Alexandria, building an advanced, deep learning, neural network-based approach to modeling news analysis. Earlier generation, simple dictionary-based approaches to modeling news analysis tend to be less robust out of sample.”
Northfield risk models with Risk Systems That Read are available from Northfield directly or through its partners, such as FactSet, Charles River Development, Thomson Reuters, and State Street.