Transformer-Based Models for Commodity Trading Price Forecasting

Transformer-based models have gained significant traction across various fields such as natural language processing (NLP), large language models (LLM), smart transportation, and finance due to their ability to capture long-term dependencies and sequence patterns. However, their application in forecasting stock or commodity trading returns remains underexplored. In this study, we aim to compare the performance of three transformer-based models—Informer, Autoformer, and Transformer with Positional Encoder unit —alongside an LSTM for forecasting commodity returns. Specifically, the two datasets used will consist of historical commodity prices, including six types of metals. These models, known for their capacity to handle long sequence data, have shown promise in financial forecasting, particularly. Still, their potential in commodity trading price forecasting has been less studied. By evaluating these models based on historical metal prices, this research seeks to determine the most accurate model for predicting commodity returns, providing valuable insights for the commodity trading and financial sectors.