Using Machine Learning and Agent-Based Simulation to Predict Learner Progress for the South African High School Education System

Aug 5, 2026

Upload by Sage

The South African high school education system faces significant barriers, including high dropout rates and uneven learning outcomes, requiring analytical approaches to diagnose and mitigate these challenges. This study integrates machine learning with agent-based modelling to simulate learner progression in public schools. Using 2019 General Household Survey data, factor analysis derives key learner characteristics, which train an XGBoost model embedded in an agent framework for Grades 8 to 12 progression. Validation against LURITS data achieved an RMSE of 2.95%, supporting effective predictive performance and practical decision support.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages133

Reading Time5 h 35 min

AccessFree to download

1 Introduction

1.1 Basic education in South Africa

1.2 The state of South Africa’s basic education system

1.3 Problem description

1.4 Research question

1.5 Methodology

1.6 Conclusion: Introduction

2 Literature review

2.1 Theoretical frameworks and key concepts

2.2 General application of techniques in education

2.3 Notew

Using machine learning and agent-based simulation to predict learner progress for the South African high school education system by Maymarie van den Heever Thesis presented in partial fulfilment of the requirements for the degree of Master of Engineering (Industrial Engineering) in the Faculty of Engineering at Stellenbosch University Supervisor: Dr Lieschen Venter Co-supervisor: Prof James Bekker December 2024 Dedicated to my Maker, Helper and Friend. Declaration By submitting this thesis electronically, I declare that the entirety of the work contained therein is my own, original work; that I am the sole author thereof (save to the extent explicitly otherwise stated); that reproduction and publication thereof by Stellenbosch University will not infringe any third-party rights, and that I have not previously in its entirety or in part submitted it for obtaining any qualification. Date: December 2024 Copyright © 2024 Stellenbosch University All rights reserved Abstract The South African high school education system faces numerous challenges, including high dropout rates and unequal educational outcomes, calling for innovative methods to analyse and address these problems. This study employs an integrated approach that merges machine learning and agent-based modelling to simulate learner progression in public high schools, illuminating the critical factors that influence educational outcomes. Using data from the 2019 General Household Survey in South Africa, factor analysis is first conducted to identify and quantify the principal characteristics defining learners. These features then train an XGBoost machine learning model, which is integrated within an agent-based framework to simulate learner progression from Grades 8 to Grade 12 . Validating the model against the Learner Unit Record Information and Tracking System dataset resulted in a root square error of 2.95%, which is indicative of the model’s ability to predict learner progression. Overall, the model represents a significant advancement in the field of educational simulation, serving as a practical tool for schools to analyse and improve learner outcomes through analytical decision-making. Opsomming Die Suid-Afrikaanse hoërskoolonderwysstelsel staar talle uitdagings in die gesig, insluitend hoë uitvalsyfers en ongelyke onderwysuitkomste, wat vra vir innoverende metodes om hierdie probleme te ontleed en aan te spreek. Hierdie studie gebruik ’n geïntegreerde benadering wat masjienleer en agent-gebaseerde modellering saamsmelt om leerdervordering in publieke hoërskole te simuleer. Deur gebruik te maak van data van die 2019 Algemene Huishoudelike Opname in Suid-Afrika, word faktorontleding eers gedoen om die hoofkenmerke wat leerders definieer, te identifiseer. Hierdie faktore word gebruik om’n XGBoostmasjienleermodel op te lei, wat geïntegreer word binne ’n agent-gebaseerde raamwerkom leerdervordering van Graad 8 tot Graad 12 te simuleer. Die validering van die model teen die LURITS-datastel het gelei tot ’n 2.95% wortel van gemiddeldekwadraatfout, wat’n aanduiding is van die model se doeltreffende vermoë om leerdervordering te voorspel. Ten slot som lewer die model ’n beduidende bydrae tot die gebied van opvoedkundige simulasie deur te dien as ’n praktiese hulpmiddel vir skole om leerderuitkomste te ontleeden te verbeter deur analitiese besluitneming. Acknowledgements God, my Father, thank you. My gratitude exceeds, abounds, far more than I could ever think or ask. Apart from You, I can do nothing. Prof James Bekker and Dr Lieschen Venter, dare I say, we make a mean team? • Dr Lieschen, it was a pleasure working with you. I admire your pursuit of excellence and your consistent, confident, grit. Thank you for instilling discipline and determination in your students. • Prof James, while not all of our meetings were succinct, you certainly are. Not once did I have to wait for feedback—something only a handful of students can say

FAQs

What problem does this thesis address in South African high schools?

It targets high dropout rates and unequal educational outcomes that hinder effective learning progression and make educational diagnosis difficult.

How is machine learning used in the proposed approach?

Factor analysis extracts principal learner characteristics from 2019 survey data, and an XGBoost model is trained on these features to support progression prediction.

How is the prediction model validated and with what accuracy?

The model is validated using the LURITS dataset, producing an RMSE of 2.95%, indicating the ability to predict learner progression.

Aug 5, 2026

Upload by Sage