Researcher:  Merriam Thoka, University of Limpopo
Supervisor: Dr Hairong Bau, University of the Witwatersrand, Johannesburg

Tree-based machine learning (ML) models are non-linear predictive models utilized today due to their accuracy and efficiency, but understanding their decisions has received very little attention.  Recently, banks are adopting ML to compute credit score because utilizing ML or AI in credit scoring is attentive to real-time signs of a potential borrower’s creditworthiness.  The goal of the study is to create an interpretable credit scoring model that borrowers and banks can utilize to anticipate if a lendee will be able of paying back their debt, as well as to comprehend the logic behind the model’s prediction.

Researcher:  Percival Shimange, University of the Witwatersrand, Johannesburg
Supervisor: Dr R Maluta, University of Venda

Our environment’s climate change has effects on energy consumption, generation, systems, and infrastructure. In the last decade, the demand for alternative energy conversion and storage devices has increased significantly.  This project uses organic solar cells from the Harvard Clean Energy Project to predict energy band gaps using machine learning models to accelerate power supply. The HOMO-LUMO GAP was predicted using a machine learning regression model trained on HOMO, LUMO, power conversion efficiency, open circuit potential, and short circuit density.

Researcher: Malahlane Komane, University of Limpopo
Supervisor: Dr Alphonce Bere, University of Venda

The study uses a recursive partitioning approach. To investigate the age at first marriage for women living in South Africa. There are well known methods of recursive partitioning which includes CART and random forest. The aim of the research is to construct a discrete survival tree for the identification of determinants of age at first marriage for SA women.

Researcher:  Lulamile Mtebeni, University of the Witwatersrand, Johannesburg
Supervisor: Prof. Rod Alence, University of the Witwatersrand, Johannesburg

This research analyses the assumption that the root cause of terrorism is poverty and other forms of socioeconomic deprivation. By analyzing all terrorist incidents between 1997 to 2020, this research sought to contribute further to a field of study that has already enjoyed its fair share of consideration from policymakers, scholars, and the general public. Through a sequence of multiple regressions, the research ultimately found that the relationship between the said variable is statistically insignificant – thereby meaning that any sort of relationship between the said variables is indirect.

Researcher:  Kola Ijasan, University of the Witwatersrand, Johannesburg
Supervisor: Dr Babatunde Oluwayomi, University of the Witwatersrand, Johannesburg

Increase in Land Value – conventional urban land economics theories argue that transportation costs are a major element in land value.  The availability of transportation infrastructure such as motorways or train lines will dramatically increase access to services, jobs, and amenities.  According to Du and Mulley (2006), with businesses centred in particular areas and residences in another, decreasing transportation costs becomes a determining factor in the choice and demand for residential houses; and hence their value.
Demand and Supply – There is a two-way link between transportation infrastructure and land value.  There are two types of relationships: demand driven and supply driven. The supply-driven relationship states that the provision of additional transportation infrastructure will result in increased land value surrounding the enabled infrastructure. According to the demand-driven relationship, an increase in land value leads to the provision of additional transportation infrastructure.

Researcher:  Tshegofatso Kgomo, University of the Witwatersrand, Johannesburg
Supervisor: Prof. Rod Alence, University of the Witwatersrand, Johannesburg

High youth unemployment and low economic growth are challenges in South Africa (SA).  Entrepreneurship is considered an essential catalyst for this challenge. For entrepreneurship, entrepreneurial intentions (EIs) are pivotal because they are the first step to entrepreneurial activity. Whilst entrepreneurial support (ES) enables potential entrepreneurs to convert their EIs into new businesses.  The current project had 1 aim:   To gain insight into whether ES influences EIs in the South African context.

Researcher:  Taryn Michael, Sol Plaatje University
Supervisor:  Prof. Scott Hazelhurst, University of the Witwatersrand, Johannesburg

Genotype data helps understand how genetic variations lead to genetic diseases This data is extremely large to store and analyze efficiently Thus, the PLINK format was introduced to save storage space.  This research implemented a data parallel algorithm in Rust to convert raw genotype data into the PLINK bed and bim format to speed up processing time

Researcher:  Nerissa Muthayan, University of the Witwatersrand, Johannesburg
Supervisor:  Prof. Rod Alence, University of the Witwatersrand, Johannesburg

The principles of democracy are crumbling, as the internet is becoming more censored.  Internet shutdowns during elections are detrimental to democracy and indicate a dangerous disregard for human rights and effective governance.  This study examined the relationship between regime types and internet shutdowns during elections.

Researcher:  Mukhethwa Mulangaphuma, Sol Plaatje University
Supervisor: Dr Colin Chibaya, Sol Plaatje University

Swarm robotics is an AI field focused on applying swarm intelligence paradigms and methodologies to groups of homogeneous robots to perform difficult tasks that individuals cannot.  Social insects and other living creatures inspire swarm robotics. This research is aimed at creating a language for designing swarms of bee agents to perform preferred emergent behavior.  First, we investigate the primitive behaviors that define bee agents’ discrete actions at the individual level. Eighteen of these primitive behaviors have been identified as potential building blocks for the bee agent language. The identified primitive behaviors are then translated into computational terms. The identified primitive behaviors are then combined to form an ontology (a set of primitive behaviors and meta data describing how and when primitive behaviors are utilized).  Permutations of a collection of primitive actions specify the search space for the best performer set of rules for specific assignments.

Researcher:  Boineelo Sekori, Sol Plaatje University
Supervisors:  Dr Albert Whata, Sol Plaatje University

This study aims to automate karyotyping to successfully separate the overlapping human chromosomes.  The objectives are as follows:
1. Automate semantic segmentation task for separating overlapping chromosomes.
2. Perform human karyotype chromosome segmentation using deep learning algorithms.
3. Assessing deep learning models in the semantic segmentation of separating overlapping chromosomes.
4. Evaluate the performance of the adapted and modified deep learning architecture.