Abstract
This is an analysis of South Africa’s (SA) sovereign credit rating (SCR) using Naïve Bayes a Machine learning (ML) technique. Quarterly data from 1999 to 2018 of macroeconomic variables and categorical SCRs were analysed and classified to predict and compare variables used in assigning SCRs. A sovereign credit rating (SCR) is a measurement of a sovereign government’s ability to meet its financial debt obligations. The differences by Credit Rating Agencies (CRA) on rating grades on similar firms and sovereigns have raised questions on which elements truly determine credit ratings. Sovereign ratings were split into two (2) categories that is less stable and more stable. Through data cross-validation for supervised learning, the study compared variables used in assessing sovereign rating by the major rating agencies namely Fitch, Moody’s and Standard and Poor’s. Cross- validation splits the dataset into train set and test set. The research applied cross-validation to reduce the effects of overfitting on the Naïve Bayes Classification model. Naïve Bayes Classification is a Machine learning algorithm, that utilizes the Bayes theorem in classification of objects by following a probabilistic approach. All variables in the data were split in the ratio of 80:20 for the train set and test set respectively. Naïve Bayes managed to classify the given variables using the two SCR categories that is more stable and less stable. Variables classified under more stable indicates that ratings are high or favorable and those for less stable show unfavorable or low ratings. The findings show that CRAs use different macroeconomic variables to assess and assign sovereign ratings. Household debt to disposable income, exchange rates and inflation were the most important variables for estimating and classifying ratings.