Collision Course: Challenges with Road Traffic Accident Data in South Africa

Abstract

Effective data driven decision making requires sufficient evidence and skill. Sufficient evidence is sometimes measured through the availability of enough data and availability of enough signal in that data. That is, the data must contain significant attributes that contribute to the occurrence or nonoccurrence of an event. Sufficient skill can be measured by the capability of decision makers, or of their subordinates, to be able to analyse and model using this data. For governments the challenge tends to lie in both spheres, having sufficient evidence and skill. In this paper, we use road traffic accident (RTA) data to discuss some of these challenges and investigate modelling using data from road traffic accidents in South Africa. Firstly, we will look at how RTA data is collected and discuss some of the challenges in this way of handling data. Second, we investigate reasons for collecting this data, discuss current research to mitigate RTAs in South Africa; and third, we present results from the analysis and modelling of RTAs using a dataset from Soshanguve, South Africa. After performing exploratory data analysis to summarize the main characteristics within the data, we saw a need to investigate the impact that heavy duty licences (codes: $\boldsymbol{C}1, \boldsymbol{EC}1$ and $\boldsymbol{EC}$ ) have on road safety in South Africa. Before investigating this impact, we first segmented the data using clustering as a preliminary step to find associations within the variables in the data. The results from the analysis and modelling showed that, even though a large portion of motorists in Soshanguve, South Africa are in possession of heavy duty licences, the heavy duty motorists drive motor cars/station wagons more than combis/minibuses and light delivery trucks. We also found that these motorists sustained the most serious injuries in comparison to light duty licence motorists (code: $\boldsymbol{B}$ and $\boldsymbol{EB}$ ).