Researcher:  Stuart Morrison, University of the Witwatersrand, Johannesburg
Supervisor: Prof, Rod Alence, University of the Witwatersrand, Johannesburg

In this work I seek to understand the effect that fatalities have on the accuracy of election violence datasets, by asking “what happens to the quality of the dataset if we create a dataset that includes non-fatal incidents?” The results from the study show that including non-fatal events can decrease the accuracy and overall quality of the dataset by including too much noise and that the fatalities themselves help in describing how the actors interact with each other.

Researcher: Sizo Mosibi, University of the Witwatersrand, Johannesburg
Supervisor: Prof. Yudhvir Seetharam, University of the Witwatersrand, Johannesburg

Predicting future asset prices with higher accuracy has remained a formidable undertaking. Series of techniques have been developed and implemented to attain the intended result, and significant progress has been made. The use of a sentiment index as an independent variable in a machine learning model to predict asset prices, which utilizes long short-term memory, is one of the techniques employed. We developed a sentiment index to augment the South African economy and included it into the LSTM model to estimate future Sasol prices. We discovered using a sliding window strategy that using raw price in the LSTM model beats the inclusion of investor sentiment index as a variable. We also observed that the arbitrage pricing model (APT) performed the worst at predicting asset returns because it failed to account for non-linearity in price evolution.

Researcher: Shalati Baloyi, University of the Witwatersrand, Johannesburg
Supervisor: Prof. Glen Nwaila, University of the Witwatersrand, Johannesburg

Mining companies are subject to high risk and uncertainty caused by an underestimation or overestimation of mineral resources. Over the years, various techniques have been implemented to estimate mineral resources. This research compared the effectiveness of Ordinary Kriging (OK) and Sequential Gaussian simulations (SGSIM) in the estimation and quantification of uncertainty in mineral resources. Mineral grade data was transformed and used to quantify spatial continuity using variograms. Unsampled locations were estimated using OK and SGSIM was used to quantify uncertainty. The results found suggested that SGSIM is more effective in estimating mineral grade and quantifying its uncertainty compared to OK.

Researcher: Tshepiso Rangongo, University of Pretoria
Supervisor: Renate Thiede and Dr Inger Fabris-Rotelli, University of Pretoria

Remote sensing (RS) is the process of obtaining information about something without making physical contact with it.  Remote sensing images are processed to help identify areas. This process is known as land cover image classification.  There exists many algorithms that can classify land from images, and many ways of assessing their performances.

Researcher: Beauty Gama, , University of the Witwatersrand, Johannesburg
Supervisors: Prof. Rod Alence Prof. Sumaya Laher, University of the Witwatersrand, Johannesburg

Digital transformation has become an integrated part of our lives. Minimal inclusion on the impact people’s experiences have on transformation. Digital transformation cyclically interact with technology, business & society. Digital divide still poses implementation challenges. Ecological development & modernisation captures the socio-psychological influences.

Researcher: Naledi Mogane, University of the Witwatersrand, Johannesburg
Supervisors: Dr Ritesh Ajoodha and Dr Ashwini Jadhav, University of the Witwatersrand, Johannesburg

Forecasting rainfall is a critical technique for preventing climate-related risks and ensuring long-term management. This study presents the application of neural networks to predict rainfall. To accomplish this, we used a Long Short-Term Memory (LSTM), a type of recurrent neural network (RNN), and we compared the results with the Autoregressive Integrated Moving Average (ARIMA) model. The statistical effectiveness of the models reveals that the LSTM model can predict monthly rainfall in the catchment with reasonable accuracy.

Researcher: Kudzai Sibanda, University of the Witwatersrand, Johannesburg
Supervisor: Dr Tapiwa Gundu, Sol Plaatje University

In a digital world where we put our lives on our gadgets, we deserve assurance that our data is safe from any external entities. We want our gadgets to notify us of any threat to our privacy and a way to deal with these threats. We need intrusion detection systems.

Researcher: Palesa Khoza, University of the Witwatersrand, Johannesburg
Supervisor: TBA

This study reviewed 18128 discharge records issued in the department of surgery at Chris Hani Baragwanath Academic Hospital and examined 28 fields to assess the field completion rate of electronic discharge summaries.  To evaluate the quality of information in electronic discharge summaries, a word count of three clinically significant fields was conducted.

Researcher: Emma Smith, University of the Witwatersrand, Johannesburg
Supervisor: Prof. Rod Alence , University of the Witwatersrand, Johannesburg

South Africa experiences high levels of protest across various sectors. Police response to protests is varied and can lead to scenarios where excessive force is used on seemingly small or unthreatening protests.  This research report asks the following: Which protest characteristics increase the likelihood of a protest being met with police repression?

Researcher: Andre Kleynhans, University of Pretoria
Supervisor: Prof. Ritesh Ajoodha, University of the Witwatersrand, Johannesburg

Bitcoin (BTC) return prediction is a fairly difficult task. Past literature focuses on models to ensure the best return prediction. Little focus is on discovering the best features for bitcoin return prediction.  This research report attempts to bridge this gap. This research report finds the most important features from commonly used features used in the literature.