Exploring the Effectiveness of Deep Learning Models in Forecasting Commodity Prices: A Case Study on Gold

Suga, Abhishek (2024) Exploring the Effectiveness of Deep Learning Models in Forecasting Commodity Prices: A Case Study on Gold. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (3MB) | Preview
Preview	PDF (Configuration Manual) Download (85kB) | Preview

Abstract

This research work presents how efficient deep learning models, such as the Long Short-Term Memory Network (LSTM), have proved in carrying out the forecasting of gold price with high frequency hourly time series data. Compared to other traditional methods, which usually take the input as daily or low resolution data, hourly data may provide finer temporal patterns that will yield more precise predictions.

The key features that will form the basis of the dataset are: Open, High, Low, Close, and Volume. Later, this dataset is augmented further with some external economic indicators like crude oil prices, and volumes to see how this affects the performance of the model. Two models were tested in the experiments: XGBoost is a gradient-boosted decision tree algorithm, while LSTM is a neural network model designed for sequential data. Experiments were conducted in two settings: (1) using the original gold dataset and (2) augmenting the dataset with crude oil features. To have a strict comparison between different models, several metrics such as RMSE, R2, MAE, and sMAPE were adopted. The results prove that LSTM is much better at capturing temporal dependencies and generalises well in all folds, even during volatile market conditions. However, XGBoost has shown variability, especially regarding outliers. The inclusion of crude oil features improved both models, thus confirming the integration of external economic indicators.

This research reveals the potential of high frequency data and deep learning techniques for financial forecasting, thus opening ways for more sophisticated approaches in further studies.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Shahid, Abdul UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science H Social Sciences > HG Finance > Investment Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	05 Sep 2025 11:08
Last Modified:	05 Sep 2025 11:08
URI:	https://norma.ncirl.ie/id/eprint/8819

Actions (login required)

View Item