Road traffic accidents are a major public health concern, resulting in an estimated 1.2 million deaths and 50 million injuries worldwide each year. Dubai in particular experiences a high rate of such accidents. Research on road safety has been conducted for several years, yet many issues still remain undisclosed and unsolved. Specifically, the relationships between drivers' characteristics and road accidents are not fully understood. In this work, we started by collecting a dataset between 2008 and 2010 from Dubai Police. After preprocessing, we modeled the data to 19 attributes and 5 classes. We used WEKA data mining software with the 4 classifier methods (Decision trees, Rules induction, BayesNet, and MultilayerPerceptron). We applied data mining technologies to link recorded accident, driver, and road factors to accident severity in Dubai, and generated a set of rules that could be used by the Dubai Police to improve safety. Empirical results showed that the developed models could classify accidents within reasonable accuracy. The comparison of these classifiers showed that the neural networks classifier (MultilayerPerceptron algorithm) is the best classifier for all classes. We generated recommendations and conclusions.© 2005 - 2013 JATIT & LLS. All rights reserved.