Data Science

Data collection across South Africa needs to be significantly improved to allow for better decision-making. This is according to students who participated in the Department of Science and Technology's (DST's) data science training programme.

Data Science for Impact and Decision Enhancement (DSIDE) is an intensive training programme that aims to solve real-life problems using multiple technical disciplines, including computer science, analytics, mathematics, modelling and statistics. The Council for Scientific and Industrial Research (CSIR), an entity of the DST, is responsible for implementing the programme.

On 31 January 2019, 16 teams comprising 65 students showcased their projects at the DST. The projects sought to adapt a visual analytics framework with goals that included understanding datasets through interactive visual exploration and model development. The information gathered was intended to trigger actions towards better decision-making for various users.

This was the fourth time the DST was hosting the event, which took the form of presentations and exhibitions. The event was the culmination of an intensive, 12-week, mentor-guided and learn-by-doing training programme.

The DSIDE teams applied their data to a range of real-life issues, including preventing service delivery protests, improving service delivery at clinics, boosting access to the Internet, and clearing bottlenecks in the Presidential hotline.

Zinzi Villo, a University of Cape Town postgraduate information systems student, believes that improved data collection can prevent service delivery issues from spiralling out of control. Villo and her team members were tasked with developing a model that uses machine learning to predict service delivery protests before they happen. The aim was to assist the Department of Planning, Monitoring and Evaluation (DPME) in tackling the ongoing challenges related to the performance of provincial departments, particularly in the delivery of basic services.

With raw data obtained from Statistics SA, Villo's team developed a dashboard to help identify service delivery challenges and establish an early warning system for issues such as complaints, budget appropriations and expenditure, 30-day payments and vacancy rates, among others.

The team surveyed data from the Presidential Hotline, the Department of Health, the South African Social Services Agency (SASSA) and the South African Police Service over the 2014-15 and 2016-17 periods.

"Based on the findings we made, this information can really help government to know who are the people complaining and why, where can more resources be invested, and what other interventions are needed," said Ms Villo.

Some of the findings were that calls to the Presidential Hotline had decreased by an average of 3 367 complaints per year over seven years, with the most frequent complaints relating to poverty, RDP housing, and personal documentation such as identity documents.

The data indicated that hospital cleanliness was a major issue in the Northern Cape, North West and Gauteng; that patients felt disrespected by hospital staff in Mpumalanga and the Western Cape; and that citizens felt unsafe at SASSA pay points in the Eastern Cape and Mpumalanga.

The Free State and Western Cape stood out for lack of communication between citizens and the police, while citizens in the Northern Cape reported discrimination and disrespect as major issues.

Ms Villo urged for a more reliable method of collecting data, saying that, given her experience as a citizen of the Eastern Cape, there was no way to explain a 75 percent approval of clinics in the province.

"We experienced a lot of challenges in the data we analysed. Some of it was duplicated, while some was missing totally," she said, suggesting the development of online applications and other methodologies for collecting reliable data.

Another team of students, who looked at Internet access across the country, urged government and cellphone service providers to help improve access to the Internet, particularly in underdeveloped provinces.

Mbali Khanyile, a master's student in remote sensing and geographic information systems at the University of Johannesburg, said the country could not be expected to prepare for the Fourth Industrial Revolution if more than 50 percent of people did not have reliable access to the Internet.

According to the data taken from the 2011 Census, Limpopo was worst off among South Africa's nine provinces, with 75 percent of people reporting lack of Internet access. In rural areas in particular, people often did not have access even to the older 3G networks.

The Western Cape Water Allocation and Registration Project was part of a larger study to verify and validate water use in the water-stressed Breede-Gouritz water catchment area in the province.

This project sought to determine whether customers in the Coastal, Gouritz and Olifants regions were using water lawfully or unlawfully, and to classify them by location, economic sector and type (individuals or companies).

Information populated by customer and geographic information system (GIS) data was used to design a water management tool to help water officials and managers detect the unlawful use of water, monitor usage and plan for future demand.

Team members Kemishen Ramsing, a mechanical engineering student at the University of Pretoria, Hadio Mantyi, an engineering PhD student at Wits University, and Thato Dikobo, a BSc honours student at the University of Johannesburg, collaborated on the design of the water management tool. They concluded that up to 87% of Breede-Gouritz users could be using water unlawfully.

While the findings from these projects provide useful information, the students' participation does not end with the development of tools for better decision-making, but contributes to the growth of data science skills in the country.

Strategic Research Manager at the CSIR Meraka Institute, Quentin Williams, says that since the DSIDE programme started four years ago, a total of 150 out of 212 students had obtained employment at institutions like the South African Revenue Service, Statistics South Africa, the Reserve Bank and Investec, among others.

Success stories include projects like Rock Pulse, which students developed to help detect rock falls in mines before they could happen. Another project was being implemented by Transnet to optimise container movement at ports, while a hospital scheduling system had been developed for use by a hospital in New Zealand.

Issued by the Department of Science and Technology