Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
A graph-based framework towards trading financial markets
Pretorious, J. C. 2025. A graph-based framework towards trading financial markets. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/0d922223-6289-4084-b528-f1effce6b0f8
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
Stellenbosch : Stellenbosch University
2025
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613749298331648 |
|---|---|
| access_status_str | Open Access |
| author | Pretorious, Joshua Chance |
| author2 | Nel, Stephan |
| author_browse | Nel, Stephan Pretorious, Joshua Chance |
| author_facet | Nel, Stephan Pretorious, Joshua Chance |
| author_sort | Pretorious, Joshua Chance |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Pretorious, J. C. 2025. A graph-based framework towards trading financial markets. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/0d922223-6289-4084-b528-f1effce6b0f8 |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/132492 |
| institution | Stellenbosch University (South Africa) |
| language | English |
| last_indexed | 2026-06-10T12:41:04.390Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2025 |
| publishDateRange | 2025 |
| publishDateSort | 2025 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/132492 A graph-based framework towards trading financial markets Pretorious, Joshua Chance Nel, Stephan Van Vuuren, Jan Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering. Graph theory Artificial intelligence -- Financial applications Financial markets -- Forecasting Stock exchanges -- Data processing UCTD Pretorious, J. C. 2025. A graph-based framework towards trading financial markets. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/0d922223-6289-4084-b528-f1effce6b0f8 Thesis (MEng)--Stellenbosch University, 2025. ENGLISH ABSTRACT: The term financial markets refers broadly to any marketplace in which the trading of securities occurs. The finance sector plays a notable role in any country’s economic development by providing liquidity, facilitating investment, and allocating resources efficiently. The behaviour of these markets is therefore a good indicator of a country’s economic health. The dynamic nature with respect to financial markets provides value by enabling investors to generate capital through the trading of financial securities, which in turn provides a source of funding for companies and governments. Moreover, it facilitates the transfer of risk from individuals who are risk-averse to those who are willing to assume risk, which helps promote economic growth and innovation. The volatile and non-stationary nature of financial markets can, however, also induce risk and uncertainty, which can have negative effects on the economy. Consequently, there is a need for an improved understanding of financial market behaviour so as to mitigate risk and promote economic stability. Actionable insight may be derived from financial market data if abstracted and analysed by means of an appropriate modelling approach. The application of network science to financial markets is a promising approach to this end, as it enables the abstraction of intricate interdependencies embedded between securities and the discernment of underlying structural similarities by means of a mathematical construct known as graphs. The application of various network algorithms may be employed so as to construct investment portfolios, according to which the devised strategy distributes an investor’s capital across an amalgamation of securities varying in respect of risk profile. Moreover, spatial-temporal graph neural networks represents a promising approach towards inferring insight from multivariate time series data. There are, however, various complexities involved with modelling the financial market as a network and subsequently deriving actionable insight (by means of a trading strategy) therefrom. In this thesis, a generic framework is proposed to this end — it conceptually represents an architectural pipeline that may be employed towards transforming raw historical market data to an appropriate graph-based format and subsequently performing graph-based analyses by means of network algorithms, such as community detection, centrality analysis, minimum spanning trees, and graph neural networks. The proposed framework comprises four functional components, namely a Processing component, an Asset selection component, a Forecasting component and a Trade analysis component. Each component addresses a notable step in the overarching process in respect of algorithmic trading. The practicality of the proposed framework is demonstrated via two computerised instantiations thereof. More specifically, two different data sets are considered, each of which differs in respect of the underlying financial market, i.e. forex and stocks, in terms of complexity and statistical properties. An algorithmic verification study was conducted prior to these instantiations in order to verify the functional correctness of the network algorithms considered. Finally, the insight gleaned with respect to the graph-based analyses are leveraged and implemented by means of a popular trading simulation environment. The results are then presented by means of standard performance metrics established in the trading domain and subsequently discussed. Furthermore, the framework is validated by means of four independent subject matter experts, during which the methodological utility of the proposed framework and its applicability to real-world operations is corroborated. AFRIKAANSE OPSOMMING: Die term finansi¨ele markte verwys wyd na enige markplek waarin die verhandeling van sekuriteite plaasvind. Die finansiesektor speel ’n noemenswaardige rol in enige land se ekonomiese ontwikkeling deur die verskaffing van likiditeit, die fasilitering van beleggings, en die doeltreffende toewysing van hulpbronne. Die gedrag van hierdie markte is dus ’n goeie aanduiding van ’n land se ekonomiese gesondheid. Die dinamiese aard van finansi¨ele markte lewer waarde deur beleggers in staat te stel om kapitaal te genereer deur die verhandeling van finansi¨ele sekuriteite, wat op sy beurt ’n bron van finansiering vir maatskappye en regerings verskaf. Verder vergemaklik dit die oordrag van risiko van individue wat risiko-avers is na di´e wat bereid is om risiko te aanvaar, wat ekonomiese groei en innovasie bevorder. Die onstabiele en nie-stasionˆere aard van finansi¨ele markte kan egter ook risiko en onsekerheid veroorsaak, wat negatiewe effekte op die ekonomie kan hˆe. Gevolglik is daar ’n behoefte aan ’n verbeterde begrip van finansi¨ele markgedrag om risiko te verminder en ekonomiese stabiliteit te bevorder. Insig kan afgelei word van finansi¨ele markdata indien dit gemodileer en geanaliseer word deur ’n geskikte benadering. Die toepassing van netwerkwetenskap op finansi¨ele markte is ’n belowende benadering tot hierdie doel, aangesien dit die abstraksie van ingewikkelde onderlinge afhanklikhede tussen sekuriteite moontlik maak en die analise van onderliggende strukturele ooreenkomste deur middel van ’n wiskundige abstraksie wat grafieke genoem word. Die toepassing van verskeie netwerkalgoritmes kan gebruik word om beleggingsportefeuljes saam te stel, waardeur die strategie ’n belegger se kapitaal versprei oor ’n samestelling van sekuriteite wat verskil in terme van risikoprofiel. Verder verteenwoordig ruimte-tydelike grafiek-neurale netwerke ’n belowende benadering tot die aflei van insig uit multivariate tydreeksdata. Daar is egter verskeie kompleksiteite verbonde aan die modellering van die finansi¨ele mark as ’n netwerk en die daaropvolgende afleiding van insig (deur middel van ’n handelsstrategie). In hierdie tesis word ’n generiese raamwerk voorgestel wat konseptueel ’n argitektoniese pyplyn verteenwoordig. Hierdie pyplyn kan gebruik word om rou historiese markdata om te skakel na ’n gepaste grafiekgebaseerde formaat, en om daaropvolgende grafiekgebaseerde analises uit te voer deur middel van netwerkalgoritmes soos gemeenskapsdeteksie, sentraliteitsanalise, minimum-spanningbome en grafiekneurale netwerke. Die voorgestelde raamwerk bestaan uit vier funksionele komponente, naamlik ’n Verwerkingskomponent, ’n Batekeurings-komponent, ’n Voorspellingskomponent en ’n Handelsanalise-komponent. Elke komponent adresseer ’n belangrike stap in die oorsigproses vir algoritmiese handel. Die praktiese bruikbaarheid van die voorgestelde raamwerk word gedemonstreer deur middel van twee gerekenariseerde instansiasies daarvan. Meer spesifiek, twee verskillende datastelle word oorweeg, elk van hulle verskil in terme van die onderliggende finansi¨ele mark, d.w.s. forex en aandele, in terme van kompleksiteit en statistiese eienskappe. ’n Algoritmiese verifikasiestudie is uitgevoer vir hierdie instansiasies om die funksionele korrekheid van die oorweegde netwerkalgoritmes te verifieer. Vervolgens word die insigte wat uit die grafiekgebaseerde analises verkry is, ge¨ımplementeer deur middel van ’n gewilde handelsimulasiestelsel. Die resultate word dan voorgehou deur middel van standaard prestasiewerklikheidsmaatre¨els wat in die handelsdomein gevestig is en daaropvolgend bespreek. Verder word die raamwerk gevalideer deur vier onafhanklike kenners, wat die metodologiese bruikbaarheid en toepaslikheid daarvan vir werklike operasies bevestig. Masters 2025-06-10T05:59:49Z 2025-06-10T05:59:49Z 2025-03 Thesis https://scholar.sun.ac.za/handle/10019.1/132492 en Stellenbosch University 304 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Graph theory Artificial intelligence -- Financial applications Financial markets -- Forecasting Stock exchanges -- Data processing UCTD Pretorious, Joshua Chance A graph-based framework towards trading financial markets |
| title | A graph-based framework towards trading financial markets |
| title_full | A graph-based framework towards trading financial markets |
| title_fullStr | A graph-based framework towards trading financial markets |
| title_full_unstemmed | A graph-based framework towards trading financial markets |
| title_short | A graph-based framework towards trading financial markets |
| title_sort | graph based framework towards trading financial markets |
| topic | Graph theory Artificial intelligence -- Financial applications Financial markets -- Forecasting Stock exchanges -- Data processing UCTD |
| url | https://scholar.sun.ac.za/handle/10019.1/132492 |
| work_keys_str_mv | AT pretoriousjoshuachance agraphbasedframeworktowardstradingfinancialmarkets AT pretoriousjoshuachance graphbasedframeworktowardstradingfinancialmarkets |