Machine learning and deep learning methods for exchange-traded fund trading

• Main Author: Jeremia, Alexander
• Other Authors: Wong Jia Yiing, Patricia
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Pattern recognition
• Online Access: https://hdl.handle.net/10356/167311
• Institution: Nanyang Technological University

Machine Learning and Deep Learning has been growing in the popularity in many fields in the past few years especially in the field of Exchange Traded Funds (ETF) due to its ability to analyse large amount of data and detect patterns that might not be oblivious to traders. Additionally, these methods can process and analyse vast amounts of data in real-time, helping traders to make informed decisions, react quickly to changes in the market, and reduce the risk of losses. Due to the interest in the field, many have done research on the performance of different type of Machine Learning models. In this report, we study the effectiveness of using different Machine Learning models for ETF trading. Furthermore, we will also study the effectiveness of combining more than one machine learning models to predict the price movement of an ETF. The machine learning models that were used in the research were LSTM and Markov Chain due to their ability to track time. This property was useful for analysing the ETF market because time plays an important role in market price movement. LSTM model was used to predict the exact price of ETF while Markov Chain was used to categorise certain price patterns to its predictions. Both models demonstrated good accuracy and outperformed the buy-and-hold strategy in generating more profitable trades The LSTM model achieved a MSE of 0.00083984 and generated a 7.4% profit between January 2022 and January 2023, whereas the buy-and-hold strategy only yielded a 4.62% profit. On the other hand, the Markov Chain model achieved a 58.94% accuracy rate in classifying specific patterns in price movement. When the two models were combined, the resulting algorithm produced a higher APY compared to using the LSTM model alone or the buy-and-hold strategy.