Applied Natural Language Processing and Machine Learning in Algorithmic Trading

Olaniyan, Rapheal. 2021. Applied Natural Language Processing and Machine Learning in Algorithmic Trading. Doctoral thesis, Goldsmiths, University of London [Thesis]

No full text available
[img] Text (Applied Natural Language Processing and Machine Learning in Algorithmic Trading)
COM_thesis_OlaniyanR_2021.pdf - Accepted Version
Permissions: Administrator Access Only until 30 November 2023.
Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (2MB)

Abstract or Description

The frequent ups and downs are characteristic of the stock market. The conventional predictive models that assume that investors act rationally have not been able to capture the irregularities in the stock market. They rely mainly on fundamental data such as assets, liabilities, among others. As such, these models seem to fail to capture the stock market trends that are extremely sensitive to social, economic, and political behavioural elements. As a result, behavioural finance is embraced to attempt to correct these model shortcomings by adding some factors to help capture the sentimental contagion which may be at play in determining the stock market. Many research works have attempted to establish this relationship between emotions and the stock market but, surprisingly, findings from these works have been rather conflicting due to the generic nature of sentimental information. This thesis is therefore relevant in that it helps to clarify on the relationship between sentiments and the stock market. First, this work explores different sources of data including pre-processed sentiments and sentiments extracted directly from raw financial news data based on a proposed novel BERT-based Natural Language Processing (NLP) algorithm. Also, most of the previous studies claiming that emotions have predictive value on the stock market do so by developing various machine learning predictive models, but do not validate their claims rigorously. Such findings may be clearly misleading. This problem is addressed in this thesis with a focus on the relevance and appropriateness of model applicability and statistical validation. Finally, this thesis proposes an approach that incorporates our proposed NLP and stock market trading algorithms. The NLP algorithm automatically extracts the sentiment polarities from financial news and activates the proposed stock market trading algorithm to predict the directions of the stock market prices.

Item Type:

Thesis (Doctoral)

Identification Number (DOI):

https://doi.org/10.25602/GOLD.00030952

Keywords:

NLP; algorithmic trading; stock market; SVM, BERT; sentiment; Granger causality; technical analysis indicators, contract for difference, and machine learning

Departments, Centres and Research Units:

Computing

Date:

30 November 2021

Item ID:

30952

Date Deposited:

22 Dec 2021 10:42

Last Modified:

18 Feb 2022 17:36

URI:

https://research.gold.ac.uk/id/eprint/30952

View statistics for this item...

Edit Record Edit Record (login required)