ABSTRACT
Road traffic accidents (RTAs) are the leading cause of injuries and fatalities worldwide, especially in developing countries. Addis Ababa, with a demographic population estimated with at least 4 million dwellers, is the largest urban center in Ethiopia has the high road crash rate in the world. Recently, RTAs have become a challenging public health concern within the city as the trend of RTAs injuries and death is increasing drastically. To date, few studies have been proposed to identify the crucial factors contributing to road crashes and to classify fatalities. In addition, most AI applications lack explainability/transparency. In this paper, we developed an explainable method capable of classifying accident severity accurately and providing interpretations and explanations about the pivotal features and factors that may lead to road accidents based on the proposed graph feature selection for the sake of improving road safety.
The recorded results on Addis Ababa RTAs dataset confirm the validity of our proposed method and its superiority using five evaluation metrics and three well-known classification algorithms.
