Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem
Jacobs, R. 2025. A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem. Unpublished doctoral dissertation. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/f6550db7-f565-4b1b-98...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Stellenbosch : Stellenbosch University
2025
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867614131451854848 |
|---|---|
| access_status_str | Open Access |
| author | Jacobs, Rentia |
| author2 | Grobler, J. |
| author_browse | Grobler, J. Jacobs, Rentia |
| author_facet | Grobler, J. Jacobs, Rentia |
| author_sort | Jacobs, Rentia |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Jacobs, R. 2025. A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem. Unpublished doctoral dissertation. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/f6550db7-f565-4b1b-985b-b896b305b7fb |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/132514 |
| institution | Stellenbosch University (South Africa) |
| last_indexed | 2026-06-10T12:47:09.638Z |
| 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:10019.1/132514 A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem Jacobs, Rentia Grobler, J. Grobbelaar, S. S. Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering. Decision support systems Health services accessibility Social entrepreneurship -- Health aspects Geographic information systems UCTD Jacobs, R. 2025. A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem. Unpublished doctoral dissertation. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/f6550db7-f565-4b1b-985b-b896b305b7fb Thesis (PhD)--Stellenbosch University, 2025. ENGLISH ABSTRACT: The majority of the African population, consisting of low-to-middle-income communities, heavily relies on poorly funded, overcrowded, under-resourced, and unevenly distributed public health facilities. Moreover, most people in Africa cannot afford private medical insurance, leaving them unable to access the expensive healthcare services offered by the well-funded private healthcare sector. In recent years, healthcare social enterprises have emerged to fill the gap between the public and private healthcare sectors by providing high-quality, affordable healthcare services in marginalised communities. Expanding successful healthcare social enterprises would effectively alleviate the burden of public healthcare facilities and ultimately improve access to affordable, high-quality healthcare. The aim of expansion and replication, also known as horizontal scaling up, is to widen the market presence of innovations by replicating them in new geographical areas to serve a broader population and extend their reach to different target groups. Horizontal scaling up, however, poses a significant challenge for healthcare social enterprises. Hence, the impact of healthcare social enterprises is usually limited to a specific region or local community. Although horizontal scaling up is a multidimensional concept, it requires the strategic selection of geographical locations for healthcare social enterprises in order to expand their services and coverage in marginalised communities marked by service gaps. To reduce the complexity of horizontally scaling up healthcare social enterprises requires a computerised decision support system capable of addressing the complexities of the healthcare social enterprise location problem. A novel generic machine learning-aided multi-objective optimisation framework is proposed in this dissertation to aid the development of decision support systems capable of addressing the healthcare social enterprise location problem. The proposed framework comprises three components: the processing component, the machine learning component, and the optimisation component. The processing component is devoted to managing and storing data, creating additional data, processing data, and preparing data for the subsequent framework components. The machine learning component acts as an objective function estimator by predicting the market penetration rate of healthcare social enterprises that are yet to be established in underserved communities. The predictions obtained in the machine learning component are provided as input to the optimisation component of the framework. The optimisation component facilitates the location selection of healthcare social enterprises in low-to-middle-income communities. In order to address the healthcare social enterprise location problem, two multi-objective mathematical models are proposed for the optimisation component of the framework. The two multi-objective mathematical models proposed in this dissertation, to the best of the author’s knowledge, are the first healthcare facility location models specifically aimed at addressing the HSE location problem. Unlike existing healthcare facility location models, the proposed models incorporate the objective of maximising the market penetration rate of candidate locations, a critical factor for the sustainable location selection of healthcare social enterprises in low-to-middle-income communities. Three metaheuristics were employed to determine the best metaheuristic for solving the first model: the adaptive geometry estimation-based multi-objective evolutionary algorithm 2 (AGEMOEA2), the non-dominated sorting genetic algorithm III (NSGA-III), and the s-metric selection evolutionary multi-objective optimisation algorithm (SMS-EMOA). The empirical results indicated that AGE-MOEA2 outperformed the other algorithms for solving the first model. Moreover, for the second model, AGE-MOEA2, NSGA-III, SMS-EMOA, and strength Pareto evolutionary algorithm 2 (SPEA2) were employed. The empirical results showed that SPEA2 was the best metaheuristic for solving the second model. To the best of the author’s knowledge, this dissertation represents the first application of evolutionary algorithms from all three major categories — dominance-based, indicator-based, and decomposition-based approaches — in the context of healthcare facility location problems. A fully functional computerised decision support system instantiation of the proposed framework is demonstrated as a proof of concept. Moreover, in order to demonstrate the practical applicability of the research presented in this dissertation, the instantiation of the framework is applied to two case studies involving real-world data. The outcomes of these case studies were validated by the subject-matter expert from the industry partner attached to this dissertation. The subject-matter expert confirmed that the decision support system instantiation of the proposed framework provided credible, high-quality solutions that align with the current location decisions of the industry partner for future clinics. AFRIKAANSE OPSOMMING: Die meerderheid van die Afrika-bevolking, wat uit lae- tot middelinkomste-gemeenskappe bestaan, maak hoofsaaklik staat op onderbefondsde, oorvol, swak toegeruste en ongelyk verspreide openbare gesondheidsfasiliteite. Buitendien kan die meeste mense in Afrika nie private mediese versekering bekostig nie, wat hulle verhoed om toegang tot die duur gesondheidsdienste, wat deur die goed befondsde private gesondheidsorgsektor aangebied word, te verkry. In onlangse jare het sosiale gesondheidsorgondernemings na vore gekom om die gaping tussen die openbare en private gesondheidsorgsektore te oorbrug deur hoëgehalte, bekostigbare gesondheidsdienste in gemarginaliseerde gemeenskappe te lewer. Die uitbreiding van suksesvolle sosiale gesondheidsorgondernemings kan die druk doeltreffend op openbare gesondheidsfasiliteite verlig en gevolglik toegang tot bekostigbare, hoëgehalte gesondheidsorg verbeter. Die doel van die uitbreiding en replisering, ook bekend as horisontale skalering, is om die markteenwoordigheid van innovasies te verbreed deur dit in nuwe geografiese gebiede te repliseer. Horisontale skalering bied egter ’n groot uitdaging vir sosiale gesondheidsorgonderneming wat volhoubaar gesondheidsdienste in lae- tot middelinkomste-gemeenskappe wil lewer. Gevolglik is die impak van hierdie ondernemings gewoonlik beperk tot ’n spesifieke streek of plaaslike gemeenskap. Alhoewel horisontale skalering ’n multidimensionele konsep is, verg dit die strategiese seleksie van geografiese liggings vir sosiale gesondheidsorgondernemings. Om die kompleksiteit van horisontale skalering te verminder, vereis dit die gebruik van gerekenariseerde besluitsteunstelsel wat die sosiale gesondheidsorgondernemings-plasingsprobleem in ag kan neem. ’n Nuwe generiese masjienleer-ondersteunende multi-doelwit-optimeringsraamwerk word in hierdie proefskrif voorgestel om die besluitneming vir die ligging van sosiale gesondheidsorgondernemings in lae- tot middelinkomste-gemeenskappe te ondersteun. Die voorgestelde raamwerk bestaan uit drie komponente: die verwerkingskomponent, die masjienleerkomponent en die optimeringskomponent. Die verwerkingskomponent is daarop gemik om data te verwerk en vir die daaropvolgende komponente van die raamwerk voor te berei. Die masjienleerkomponent dien as ’n doelfunksieskatter deur die markinfiltrasiekoers van sosiale gesondheidsorgondernemings, wat nog in lae- tot middelinkomste-gemeenskappe gevestig moet word, te voorspel. Die voorspellings wat deur die masjienleerkomponent verkry word, dien as insette vir die optimeringskomponent van die raamwerk. Die optimeringskomponent fasiliteer die proses om liggings vir sosiale gesondheidsorgondernemings in lae- tot middelinkomste gemeenskappe te kies. Om die sosiale gesondheidsorgondernemings-plasingsprobleem aan te spreek, word twee multi-doelwitwiskundige modelle vir die optimeringskomponent van die raamwerk voorgestel. Volgens die beste kennis van die outeur is die twee voorgestelde modelle die eerste gesondheidsfasiliteitsplasingsmodelle wat spesifiek daarop gemik is om die plasingsprobleem van hierdie ondernemings aan te spreek. Anders as bestaande gesondheidsfasiliteitsplasingsmodelle, neem hierdie voorgestelde modelle die doelwit om die markinfiltrasiekoers van kandidaatliggings te maksimaliseer in ag. Drie metaheuristieke is gebruik om die beste metaheuristiek vir die oplossing van die eerste model te bepaal: die aanpasbare meetkundige-skatting-multi-doelwit-evolusionêre algoritme 2 (AGEseleksie evolusionêre multidoelwit-optimeringsalgoritme (SMS-EMOA). Die empiriese resultate het aangedui dat AGE-MOEA2 die ander algoritmes vir die oplossing van die eerste model oortref het. Verder is dieselfde drie algoritmes, asook die sterkte-Pareto-evolusionêre algoritme 2 (SPEA2), vir die tweede model ingespan. Die empiriese resultate het getoon dat die SPEA2 die beste metaheuristiek vir die oplossing van die tweede model was. Volgens die beste kennis van die outeur verteenwoordig hierdie proefskrif die eerste toepassing van evolusionêre algoritmes uit al drie hoofkategorieë — oorheersingsgebaseerde, aanwysergebaseerde en dekomposisie-gebaseerde benaderings — in die konteks van gesondheidsfasiliteitsplasingsprobleme. ’n Instansie van die raamwerk word as ’n bewys van die konsep gedemonstreer in die vorm van ’n volledige funksionele, gerekenariseerde besluitondersteuningstelsel. Sodoende is die raamwerk toegepas op twee gevallestudies wat ware wêrelddata gebruik om die praktiese toepaslikheid van die navorsing in hierdie proefskrif ten toon te stel. Die uitkomste van hierdie gevallestudies is deur ’n vakspesialis van die industrievennoot wat aan hierdie proefskrif gekoppel is, gevalideer. Die vakspesialis het bevestig dat die raamwerk-instansie geloofwaardige, hoëgehalte-oplossings verskaf het wat in lyn is met die vennoot se huidige besluite rakende die ligging van toekomstige klinieke. Doctoral 2025-06-10T10:25:14Z 2025-06-10T10:25:14Z 2025-03 Thesis https://scholar.sun.ac.za/handle/10019.1/132514 Stellenbosch University xxvi, 219 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Decision support systems Health services accessibility Social entrepreneurship -- Health aspects Geographic information systems UCTD Jacobs, Rentia A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem |
| title | A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem |
| title_full | A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem |
| title_fullStr | A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem |
| title_full_unstemmed | A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem |
| title_short | A machine learning-aided multi-objective optimisation decision support framework for the healthcare social enterprise location problem |
| title_sort | machine learning aided multi objective optimisation decision support framework for the healthcare social enterprise location problem |
| topic | Decision support systems Health services accessibility Social entrepreneurship -- Health aspects Geographic information systems UCTD |
| url | https://scholar.sun.ac.za/handle/10019.1/132514 |
| work_keys_str_mv | AT jacobsrentia amachinelearningaidedmultiobjectiveoptimisationdecisionsupportframeworkforthehealthcaresocialenterpriselocationproblem AT jacobsrentia machinelearningaidedmultiobjectiveoptimisationdecisionsupportframeworkforthehealthcaresocialenterpriselocationproblem |