Supervision

Potential postgraduate students looking for a supervisor can read through my general research interests as well as completed and current projects listed below. If you find my area of research interesting, then you are welcome to contact me to discuss ideas and possible project topics. I am also open to discussing other topics of interest to you in the broad area of computational intelligence.

In the Department of Decision Sciences at UNISA, we offer MSc and PhD degrees in Operations Research. You are welcome to contact me via email for more information on postgraduate studies in our department. I can also supervise postgraduate degrees in Computer Science under the School of Computing, UNISA.

General research interests

My research is in the broad area of computational intelligence (CI). This is the branch of AI that includes evolutionary computation and neural networks. The focus of my reasearch is on analysing and understanding existing CI approaches to solving optimisation and machine learning problems. Although the algorithms / models are generally easy to explain, the resulting behaviour is largely unpredictable, due to not only the stochastic nature of CI algorithms, but also due to the complex interactions between components of the system. For example, although it is easy to explain how a neural network is trained, it is not easy to know whether the resulting model will generalise well on unseen data or when it might fail at a learning task. The aim of my research is to understand the features of problem instances that make them hard to solve by algorithmic approaches and then to develop techniques that can automatically select the most appropriate algorithmic approach to solving a new problem guided by the features of the problem. In the context of optimisation, this is known as automated algorithm selection and in the context of machine learning, this is known as metalearning.

Completed supervision

  • Elizabeth Booi (with Dr MT MaseTshaba as co-supervisor), MSc in OR (UNISA), Investigating the influence of first-year expectations and experiences on student academic performance (with distinction), 2024.
  • Sibusiso Mtiyane (with Prof Mardi Jankowitz as co-supervisor), MSc in OR (UNISA), Exploring the accuracy-explainability trade-off on credit scoring classifiers (with distinction), 2024.
  • Azwidohwi Nange (co-supervisor with Prof Chris Swanepoel), MSc in OR (UNISA), An investigation into methods of hyperparameter tuning for machine learning based on two case studies, 2024.
  • Johann Pienaar (co-supervisor with Dr Anna Bosman), MIT Big Data Science (UP), Hilbert curves for efficient and informative neighbourhood sampling for exploratory landscape analysis (with distinction), 2024
  • Werner Mostert (co-supervisor with Prof Andries Engelbrecht), PhD in CS (SUN), Landscape Aware Algorithm Selection for Feature Selection, 2023.
  • Tshidiso Mazibuko, MSc in OR (UNISA), Performance improvements in machine learning approaches for fault detection and soft sensing in the process industry (with distinction), 2022.
  • Jarrod Moses (co-supervisor with Dr Anna Bosman), MIT (UP), Fitness landscape analysis of generative adversarial networks (with distinction), 2021.
  • Phlippie Bosman (co-supervisor with Prof Andries Engelbrecht), MSc in CS (UP), The Influence of Fitness Landscape Characteristics on the Search Behaviour of Particle Swarm Optimisers (with distinction), 2019.
  • Odirile Moja, MIT (UP), Ethics and Benefits of Automated Decision Making for Optimal Retirement Portfolio Risk Matching in South Africa (with distinction), 2018.
  • Christina Kroeze, MSc in CS (UP), Touch Biometrics for Unobtrusive Continuous Authentication on Smartphones (with distinction), 2015.
  • Darren Pillay, MIT (UP), The Development of an Expert System Knowledge Base for the Calibration of Stand-Alone Engine Control Units, 2015.
  • Jhani de Bruin (with Prof Jan Eloff as co-supervisor), MSc in CS (UP), Automated Usability Analysis and Visualisation of Eye Tracking Data (with distinction), 2014.

Current supervision

Doctoral students:
  • Elizabeth Booi, PhD in OR (UNISA), Responsible AI framework for implementing learning analytics in higher education.
  • Marc Gagiano, PhD in OR (UNISA), Data-driven automated convolutional neural network model selection for image classification
  • Tshidiso Mazibuko, PhD in OR (UNISA), Metalearning for unsupervised anomaly detection.
  • Evans Riba (with Dr Edinah Mudimu as co-supervisor), PhD in OR (UNISA), An investigation into univariate time series forecasting using seasonal specialists and meta-learning.
  • Darren Roos (co-supervisor with Prof George Davie), PhD in OR (UNISA), Advancing contextualised topic models through fine-tuning, alternative embeddings and performance complementarity.
Masters students:
  • Charles Britz (co-supervisor with Dr Anna Bosman), MSc in CS (UP), Loss Landscape-Aware Neural Network Training
  • Pieter Ferreira, MSc in OR (UNISA), Modelling economically relevant South African fruit fies: A data-driven approach.
  • Cameron Kromker, MSc in OR (UNISA), Modelling the impact of climate on the growth of Listeria monocytogenes in the food chain.
  • Martina Marx, MSc in OR (UNISA), Data preprocessing for effective automated synthetic speech detection.
  • Tumelo Molwantwa (with Dr Diteboho Xaba as co-supervisor), MCom in QM (UNISA), Optimising Time-to-Delivery for Grocery Delivery Systems in South Africa.
  • Christiaan Opperman (co-supervisor with Dr Anna Bosman), MSc in CS (UP), The Effects of Regularisation Techniques on Fitness Landscapes of Neural Networks.
  • Ruban Pillay, MSc in OR (UNISA), Automated algorithm selection for clustering.
  • Mokgadi Seshoka (with Dr Diteboho Xaba as co-supervisor), MSc in OR (UNISA), Understanding household electrical consumption in South Africa through data analysis and predictive modelling.
  • Tiara Singh (with Prof Chris Swanepoel as co-supervisor), MSc in OR (UNISA), A Time Series Analysis on Financial Statement Fraud in South Africa.
  • Henri van der Grijp (co-supervisor with Dr Anna Bosman), MSc in CS (UP), Using weight oscillation for driving gradient-based neural network training.
  • Francois van der Merwe (co-supervisor with Dr Anna Bosman), MSc in CS (UP), Loss Landscape Analysis for Neural Network Compression.