We reflect on the good, the bad and the ugly of the fatal accidents occurring on South Africa’s roads. The cost of human lives indisputably equates to ‘the ugly’ and the economic cost of accidents associates with ‘the bad’. ‘The good’ relates to the reduction of both these costs that may result from the entrance of self-driving cars into the South African market as well as awareness campaigns like the Arrive Alive National Road Safety Strategy. The general contribution of this paper is to raise awareness of the effects of accidents, more specifically fatal accidents. Current trends in terms of human factors as well as road and environmental factors involved in the fatal accidents on South African roads are summarised. This paper also serves as a preliminary investigation into possible factors influencing these accidents, which ought to be of interest to a very broad readership, more specifically those focusing on risk analysis, and certainly is of interest to any citizen of South Africa. Significance: • Awareness is raised of the effects of fatal accidents on South African roads. • Current trends in terms of human factors as well as road and environmental factors on road accidents are reflected upon. • The futuristic effect of self-driving cars is explored.

Segmentation (or partitioning) of data for the purpose of enhancing predictive modelling is a well-established practice in the banking industry. Unsupervised and supervised approaches are the two main streams of segmentation and examples exist where the application of these techniques improved the performance of predictive models. Both these streams focus, however, on a single aspect (i.e. either target separation or independent variable distribution) and combining them may deliver better results in some instances. In this paper a semi-supervised segmentation algorithm is presented, which is based on k-means clustering and which applies information value for the purpose of informing the segmentation process. Simulated data are used to identify a few key characteristics that may cause one segmentation technique to outperform another. In the empirical study the newly proposed semi-supervised segmentation algorithm outperforms both an unsupervised and a supervised segmentation technique, when compared by using the Gini coefficient as performance measure of the resulting predictive models.

We applied different modelling techniques to six data sets from different disciplines in the industry, on which predictive models can be developed, to demonstrate the benefit of segmentation in linear predictive modelling. We compared the model performance achieved on the data sets to the performance of popular non-linear modelling techniques, by first segmenting the data (using unsupervised, semi-supervised, as well as supervised methods) and then fitting a linear modelling technique. A total of eight modelling techniques was compared. We show that there is no one single modelling technique that always outperforms on the data sets. Specifically considering the direct marketing data set from a local South African bank, it is observed that gradient boosting performed the best. Depending on the characteristics of the data set, one technique may outperform another. We also show that segmenting the data benefits the performance of the linear modelling technique in the predictive modelling context on all data sets considered. Specifically, of the three segmentation methods considered, the semi-supervised segmentation appears the most promising.

A direct or indirect modelling methodology can be used to predict Loss Given Default (LGD). When using the indirect LGD methodology, two components exist, namely, the loss severity component and the probability component. Commonly used models to predict the loss severity and the probability component are the haircut- and the logistic regression models, respectively. In this article, survival analysis was proposed as an improvement to the more traditional logistic regression method. The mean squared error, bias and variance for the two methodologies were compared and it was shown that the use of survival analysis enhanced the model's predictive power. The proposed LGD methodology (using survival analysis) was applied on two simulated datasets and two retail bank datasets, and according to the results obtained it outperformed the logistic regression LGD methodology. Additional benefits included that the new methodology could allow for censoring as well as predicting probabilities over varying outcome periods.

The development of subnational credit-rating methodologies affords benefits for subnationals, the sovereign and its citizens. Trusted credit ratings facilitate access to financial markets and above-average ratings allow for the negotiation of better collateral and guarantee agreements, as well as for funding of, for example, infrastructure projects at superior (lower) interest rates. This paper develops the quantitative section of a credit-rating methodology for South African subnationals. The unique characteristics of South African data, their assembly, and the selection of dependent and independent variables for the linear-regression model chosen, are discussed. The methodology is then applied to the provincial Department of Health using linearregression modelling.