Ateneo researchers develop AI tools predicting market interest rates

Mathematicians from the Ateneo de Manila University (ADMU) have developed tools using artificial intelligence (AI) that can predict market interest rates, "invaluable for decision-makers in business and government."

“Interest rates are among the most important macroeconomic factors considered by financial market participants, both government and private entities, when making investment and policy decisions,” according to the study entitled "Deep Learning Approaches in Interest Rate Forecasting" written by Halle Megan L. Bata, Mark Jayson A. Victoria, Wynnona Chezska B. Alvarez, Elvira De Lara-Tuprio, Armin Paul D. Allado.

 “A reliable forecast of interest rates is a requisite to sound management of exposure to different types of risk including market risk, liquidity risk, and credit risk,” the researchers said.

They explained that the market interest rate refers to the cost of borrowing money or the reward for saving it.

"The interest rate changes based on supply and demand: if many people are borrowing but few are saving, rates go up; if the opposite happens, rates go down."

Interest rates are also affected by inflation since higher prices of commodities mean higher rates. 

The researchers tested two learning models: Multilayer Perceptrons (MLP) and Vanilla Generative Adversarial Network (VGAN). 

“MLP is a type of artificial neural network that passes the data through a series of cells, each of which processes the information in its own way and adds to the network’s overall understanding of the data,” the researchers said. “This method is often used for image recognition and machine translation because of its ability to find complex patterns in data.” 

“Meanwhile, VGAN actually consists of two networks: a generator that creates synthetic data, and a discriminator that evaluates the data’s authenticity. By working in opposition to each other—hence, ‘adversarial’ —the networks are able to refine and improve their analyses.”

The learning models produced reliable forecasts for one-month, three-month, six-month, and one-year Bloomberg Valuation Service or BVAL rates within the datasets used.

The mathematicians, according to the study, were successful in predicting key trends by using 16 domestic and global economic indicators, including inflation, exchange rates, and credit default swaps.

“We determined the best model for each interest rate tenor based on the graphical simulation and comparison of the actual and predicted values alongside the validation scores of the model. In general, both MLP and VGAN produced reliable forecasts of the interest rates,” the study read.

In general, the researchers said, both MLP and VGAN produced reliable forecasts of the interest rates.

Further, the researchers emphasized the role of AI deep learning models in financial decision-making. 

“Financial institutions could potentially deploy them to manage market, credit, liquidity, and other risks; and governments could also potentially use these models to optimize debt issuance strategies by reducing borrowing costs,” they said. — Vince Angelo Ferreras/BAP, GMA Integrated News