Deep Learning Approaches to Commodity Price Forecasting

This publication presents a condensed research summary based on my MSc dissertation, which investigated deep learning approaches for commodity price forecasting. The work compares classical statistical models with recurrent neural network architectures, including Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models. A key focus of the research was evaluating whether GRU architectures could improve predictive performance and training stability compared to standard LSTM approaches when modelling volatile, non-linear commodity price time series. The implementation and experimentation were carried out in MATLAB using publicly available economic data.