Harnessing LLMs for Time Series Forecasting: Developing a Swap-Based Hedging Strategy for Commodity Trading

This paper addresses a critical challenge faced by commodity trading companies in managing purchases between fixed bid and index-linked options. We investigate the use of Large Language Models (LLMs) for time series forecasting to create a swap-based hedging strategy that mitigates market volatility risks. By integrating LLM-driven forecasts with swap instruments, our approach enhances decision-making in commodity trades. Comprehensive empirical analysis and back-testing show the efficacy of LLMs in generating long-term forecasts, complementing traditional hedging methods. Additionally, we evaluate five LLMs and introduce an ensemble model that combines traditional machine learning with LLMs, providing a robust framework for risk management.

Access this research paper