There are a number of models that are used for mass appraisal of properties. However, the choice of a model is predicated on a number of criteria. Notable among these is to compare models predictive accuracies relative to the property market context where appraisal is undertaken or being contemplated. This study focuses on comparing predictive accuracies of mass appraisal models with a dataset of 3494 single-family property transactions in the city of Cape Town, South Africa, from 2012-2014. Five mass appraisal models including back propagation (BP) trained artificial neural networks, multiple regression model, M5P trees, support vector machine optimise with sequential minimal optimisation and additive nonparametric regression were used for the simulations. Waikato Environment for Knowledge Analysis (WEKA) explorer; an open source data mining software was used to pre-processed property data to normalised values and model property prices. The analysis shows that BP trained artificial neural networks (BP-ANNs) and M5P trees utilised in this study predicted better results with root mean squared error and mean absolute error within the acceptable threshold of 5%. But M5P trees demonstrate distinctiveness in predicted results between normalised and absolute values which require further examination. The other three mass appraisal models including multiple regression model, additive nonparametric regression and support vector machines with sequential minimal optimisation predicted results with RMSE that are higher than 5% acceptable threshold. Furthermore, contextual application of results with other studies reveal that BP-ANNs and M5P trees do not have power of universal acceptability because of varied results in other context. Therefore these models are particularly relevant to mortgage lenders, valuation offices, etc. in South Africa, but should the scope be extended to other context, application should be based on the property market features.