Researchers have developed and demonstrated an artificial intelligence (AI) program that allows them to achieve specific investment risk and return objectives for large-scale portfolios containing hundreds of assets.
“We wanted to know if we could use machine learning to improve the Sharpe ratio to get better insight into what to buy, sell or hold in your portfolio to improve your portfolio’s performance over 6- to 12-month periods,” he says . Mehmet Caner, co-author of an article on the work. “This work shows that we can.” Caner is the Thurman-Raytheon Distinguished Professor of Economics at NC State’s Poole College of Management.
The Sharpe ratio is a way of measuring the trade-off an investor’s portfolio makes between the magnitude of its returns and the risk that its holdings will lose value. It is a well-established metric used in the investment industry.
However, things get complicated when a portfolio contains hundreds of holdings, because it becomes increasingly difficult to perform risk/reward analyzes and make management decisions for all holdings.
To better manage these assets, the financial industry has increasingly turned to AI programs that use machine learning to make portfolio decisions.
Caner previously helped develop an AI program that relies on a new mathematical theorem to inform financial decision-making. However, Caner wanted to see if he could improve this AI program by incorporating a number of financial factors that the previous model didn’t take into account.
“Managing a portfolio containing hundreds of assets is a challenge,” says Caner. “It can contain a variety of stocks and commodities, most of which are related to each other in some way. How do you handle such a complicated dynamic array? We set out to train an AI program to account for a wide variety of factors with the ultimate goal of achieving a specific Sharpe ratio, and we did.
“It’s important to note that there is no ‘correct’ Sharpe Ratio; it will vary based on how much risk an investor is comfortable with. But we’ve been able to train our AI to achieve whatever Sharpe Ratio target you’ve set for in your portfolio, over the course of 6-12 months. We’ve demonstrated this in both simulations and real-world practice.”
The paper, “Sharpe ratio analysis in high dimensions: nodewise regression based on residuals in factorial models”, is published in the Journal of Econometrics. The document was co-authored by Marcelo Medeiros of the Pontifical Catholic University of Rio de Janeiro; and Gabriel FR Vasconcelos of BOCOM BBM Bank in Brazil.