Key Points
- Ateneo researchers developed MLP and VGAN deep learning models to predict money market interest rates.
- MLP offers efficient analysis with simpler structures, while VGAN excels in complex scenarios with larger datasets.
- The models incorporated 16 economic factors, including inflation and exchange rates, for robust forecasting.
- Financial institutions can manage risks, while governments can optimize borrowing strategies using these models.
Ateneo de Manila University mathematicians have created advanced artificial intelligence (AI) deep learning models to predict money market interest rates. These tools, designed to assist businesses and governments, offer a reliable way to anticipate economic shifts and make informed decisions.
Interest rates represent the cost of borrowing money or the reward for saving it, influenced by supply, demand, inflation, and central bank policies. Accurate predictions of these rates are crucial for managing economic risks and crafting sound policies. “Interest rates are among the most important macroeconomic factors considered by both government and private entities when making investment and policy decisions,” the researchers emphasized.
The team tested two deep learning models: Multi-layer Perceptrons (MLP) and Vanilla Generative Adversarial Networks (VGAN). Both models demonstrated strong predictive capabilities for Philippine Benchmark Valuation (BVAL) rates during and after the pandemic. These BVAL rates are essential indicators for short-term borrowing and investment decisions.
MLP, a neural network model, processes data through interconnected layers, identifying patterns to improve understanding. Its versatility is often applied in areas like image recognition and language translation. On the other hand, VGAN employs two networks—one generating synthetic data and the other assessing its authenticity. This adversarial setup allows VGAN to refine its predictions, particularly in complex economic scenarios.
The study revealed that both models effectively forecasted one-month, three-month, six-month, and one-year BVAL rates. MLP proved efficient for simpler analyses, requiring fewer variables while delivering reliable results. VGAN, with its sophisticated structure, excelled in handling larger datasets and complex scenarios. To ensure robust forecasting, the researchers incorporated 16 domestic and global economic indicators, including inflation rates, exchange rates, and credit default swaps.
These findings have significant implications. Financial institutions can leverage these models to manage risks such as market volatility, credit exposure, and liquidity challenges. Governments could optimize debt issuance strategies, potentially reducing borrowing costs and improving fiscal management.
The study underscores AI’s growing financial role, providing tools for more accurate economic forecasting. Published in the AIP Conference Proceedings, it highlights the need for continued innovation in neural network designs to enhance predictive accuracy and ensure widespread adoption by businesses and policymakers.
