Singapore/US dollar exchange rate forecasting using transformers

• Main Author: Zheng, Peirong
• Other Authors: Wang Lipo
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence; Business::Finance::Foreign exchange
• Online Access: https://hdl.handle.net/10356/171540
• Institution: Nanyang Technological University

The dissertation presents a detailed survey of methodologies utilizing Transformer models for the purpose of predicting currency exchange rates. Time series data is a crucial component in multiple scientific and engineering fields, and it includes exchange rates as a key form of data. This study presents three machine learning models based on the Transformer architecture, which are utilized for accurate time series forecasting. The methodology being presented utilizes self-attention processes to effectively capture complex periodicities, dynamics, and representational patterns present in the data. Furthermore, the model enables the prediction of many variables, allowing for the forecasting of exchange rates for a wide range of currencies. Our analysis highlights the impressive prediction powers of the Transformer-based model in relation to the SGD/USD exchange rate.