Kody Anand Jaxon Mistry
Developing classification and regression based machine learning algorithms for bank loan approval and stock market prediction problems
Mistry, Kody Anand Jaxon
Authors
Contributors
Bappaditya Mandal
Supervisor
Abstract
Developing machine learning (ML) algorithms to predict unseen data based on the trends of existing data has always been very challenging. Its technology demand to be applied to a wide range of problems is very high and is only expected to rise. Although neural networks-based ML algorithms have made significant impacts to make predictions in both classification and regression contexts, significant challenges still exist. One of the challenges faced while designing a ML architecture is its generality of application, which is developing a ML architecture suitable to use for multiple classification tasks. In this work, a novel ensemble architecture is proposed that can make classification decisions on a diverse range of classification tasks. Extensive experimental results and ablation studies show superior generality, by metric of accuracy statistics, of the proposed architecture over the existing classification methodologies for multiple classification tasks.
A major portion of this thesis is devoted for developing a long short term memory, recurrent neural networks (RNN) based framework that can produce regression models to predict future stock prices based off of the stock price history. The proposal of an innovative wide-long short-term memory (W-LSTM) architecture takes an RNN modified to boost performance and generality. W-LSTM architecture can handle long-term and short-term stock market prediction data very efficiently, and exhibits high performance across multiple time series datasets differing in length. Extensive experimental results and ablation studies on large datasets from 5 large companies involving long-run and short-run stock market data show the superiority of the proposed approach over the existing methodologies. This is reflected in evaluation metrics whereby a higher R2 statistic is achieved by the proposed architecture compared with existing technologies. Our model and its optimisation can be used by asset managers and day traders within the financial sector relying on being able to make accurate predictions of market behaviour. This is in order to facilitate decision making with regards to buying, selling or holding stocks and options.
Citation
Mistry, K. A. J. Developing classification and regression based machine learning algorithms for bank loan approval and stock market prediction problems. (Thesis). Keele University. Retrieved from https://keele-repository.worktribe.com/output/956367
Thesis Type | Thesis |
---|---|
Deposit Date | Oct 28, 2024 |
Publicly Available Date | Oct 28, 2024 |
Public URL | https://keele-repository.worktribe.com/output/956367 |
Award Date | 2024-10 |
Files
MistryMPhil2024
(4.3 Mb)
PDF
Downloadable Citations
About Keele Repository
Administrator e-mail: research.openaccess@keele.ac.uk
This application uses the following open-source libraries:
SheetJS Community Edition
Apache License Version 2.0 (http://www.apache.org/licenses/)
PDF.js
Apache License Version 2.0 (http://www.apache.org/licenses/)
Font Awesome
SIL OFL 1.1 (http://scripts.sil.org/OFL)
MIT License (http://opensource.org/licenses/mit-license.html)
CC BY 3.0 ( http://creativecommons.org/licenses/by/3.0/)
Powered by Worktribe © 2024
Advanced Search