Stock market prediction using deep learning approach

Citation

Ang, John Syin and Ng, Kok Why and Chua, Fang Fang (2022) Stock market prediction using deep learning approach. Journal of Engineering Science and Technology., 17 (5). pp. 3174-3186. ISSN 1823-4690

[img] Text
4.pdf
Restricted to Repository staff only

Download (391kB)

Abstract

LOB tracks the outstanding limit order for a stock or other security. LOB data is often used as an input for high-frequency trading and price prediction. Recent studies have shown that the DL methods are superior to the ML methods in stock prediction because they can extract important features for prediction. There is no existing study on stock prediction that profiles stocks. This can be overgeneralised as it predicts each stock equally when they have similar historical price movements. Moreover, the existing high-frequency stock price prediction models do not consider learning the long-term information. The objective of this study is to find out whether profiling the stocks and extending the length of the input data can further improve the prediction accuracy. The method used for this study is ESET-ConvNet, a method, which is proposed to combine two existing models, TCN and SENet with an additional embedding layer to learn the embedding of each stock. The result shows that profiling stocks and extending the input size led to an improvement in accuracy, with a trade-off in computation time. This study conclude that the main contribution of this study is a model architecture with better accuracy in stock prediction problem.

Item Type: Article
Uncontrolled Keywords: Deep learning, High frequency trading, Limit order book modelling, Stock prediction, Time series classification
Subjects: Q Science > Q Science (General) > Q300-390 Cybernetics
Divisions: Faculty of Computing and Informatics (FCI)
Depositing User: Ms Nurul Iqtiani Ahmad
Date Deposited: 01 Dec 2022 03:04
Last Modified: 01 Dec 2022 03:04
URII: http://shdl.mmu.edu.my/id/eprint/10872

Downloads

Downloads per month over past year

View ItemEdit (login required)