agsandrew - Fotolia
The U.K.'s Brexit vote was something few prognosticators saw coming prior to the June 23 referendum.
But once the results were in, it was clear the vote to leave the European Union would have a major impact on financial markets. The pound sterling fell in value by 11% two days after the vote, and both the Dow Jones Industrial Average and the London Stock Exchange's FTSE 100 index lost more than 2% of their total value. This left millions of traders all over the world scrambling to find safer investment positions.
But at least one group of investors was relatively calm, according to Omer Cedar, CEO of Omega Point Research Inc., a New York-based software company that sells analytics tools to help investment managers review their portfolios for risks. Omega Point's customers saw little change in the value of their investments after the Brexit vote, Cedar said, giving most of the credit to his company's Spark-based machine learning platform.
The root of the investing success during this tumultuous time was a data-driven strategy fueled by machine learning models that take into account more than 50 economic indicators, including company-specific reports and broader macroeconomic data. The models are part of the Omega Point Portfolio Intelligence Platform, which is built on Databricks' cloud-based Apache Spark distribution. The Omega Point software utilizes a combination of homegrown machine learning algorithms and prebuilt algorithms from MLlib, Spark's machine learning library.
Fund administrators, who are the primary users of the software, are able to review potential risks to the portfolios they manage and test different trading strategies to see how various options would affect investment risks. The machine learning platform assesses characteristics of a high-risk environment over time and quantifies that risk for investors. It is sold as a service and customers pay based on how many users need to have access.
There's no shortage of analytics tools available for supporting this kind of quantitative investing. Omega Point is among a small number of vendors offering advanced portfolio intelligence tools aimed at investment managers -- another example is Novus Partners, also based in New York. There also are portfolio managers that specialize entirely in data-driven investing strategies. In fact, the investing community was quicker to pick up on the potential benefits of using data to drive decisions than most other industries, a trend that initially gained some steam in the late 1990s and early 2000s.
The key to making this approach work is being able to look at data from many sources, Cedar said. That gives a fuller picture of the investing environment and picks up on signals that might be missed by predictive models only looking at one or two data sources. Spark makes sense for such a platform, he said, because it can interface with numerous data sources and keep data in-memory, which speeds the time it takes to get meaningful insights from analytics applications.
"Anyone can build one or two characteristics [into a model], but to understand how everything correlates together is more complicated," Cedar said. "You're able to pull those [sources] in and have Spark run a massive amount of distributed processes. You're able to do it in real time, interactively."
As sophisticated as the machine learning platform is, Cedar said, it did not predict the outcome of the Brexit vote itself, nor will it predict the next world event that shocks financial markets. Instead, he explained, the Omega Point technology seeks to predict only the market risks that may accompany events.
Keep them straight: Machine learning and statistics are not the same thing
Beginners get a crash source in machine learning development
AI development driven by machine learning