Workshop: Applying Machine Learning to the Financial Sector

Last week, Bloomberg was honored to partner with the Data Science Institute at Columbia University to host the “Machine Learning in Finance” workshop. This daylong event engaged students and professionals directly with real world examples of how machine learning is impacting the way professionals interact with the market, from risk management to regulation.

Stefano Pasquali, global head of liquidity research for Bloomberg Enterprise Solutions, was joined by Stephen Purpura (Context Relevant), Terrence Hendershott (Haas School of Business, UC Berkeley), Scott Bauguess (SEC), Tobias Preis (Warwick Business School), Shawn Mankad (Robert H. Smith School of Business), William Morokoff (Standard and Poors) and Marti Subrahmanyam (NYU Stern) to present case studies and lessons to a full auditorium audience.

Pasquali talked about liquidity assessment and how machine learning can assist the financial industry in the evaluation of assets. Currently, there is no industry-standard definition of liquidity, and it’s difficult to estimate due to lack of data and transparency. Bloomberg developed a proxy, using machine learning, which addresses market demands using Bloomberg data, analytics and pricing. Our tool can show the probability of selling a specific volume at a specific price, the expected cost of liquidation and maximum volume, and the expected days to liquidate a specific volume (given a maximum market impact.)

“Every market participant and regulator needs to have consistent liquidity measurement across different asset classes,” said Pasquali. “The core challenges,” he continued, “are lack of data and inconsistent methods of measurement. The final output should be data-driven in order to discover patterns in the financial information. At the same time it must also account for uncertainty.”

Walking through examples ranging from small to large companies across the globe, Pasquali demonstrated how a combination of models using machine learning can help predict liquidity with consistent success. However, he noted that U.S. market estimates today are able to be more liquid due to transparency and the accessibility of data which allows the application of machine learning to have a bigger impact.

The enthusiastic attendance and engagement during Pasquali’s presentation were solid indicators of the continued interest and adoption of this technology in the financial sector and underscores Bloomberg’s leadership in the area. Liquidity analysis is just one part of Bloomberg’s expansive application of machine learning. We look forward to sharing more as we continue to invest in this area.