Characterization of municipalities with high road traffic fatality rates using macro level data and the CART algorithm

Abstract

Road traffic accidents occur daily caused by different factors leading to varying degrees of injury severity. Considering this, many studies have been developed to identify and understand these factors to implement preventive actions. A Decision Tree (DT) is one of the techniques that can generate classifications and predictions by detecting a priori unknown patterns. This study aims to identify the characteristics of municipalities with high and very high fatality rates caused by traffic accidents, using a macro level dataset and a DT algorithm (CART - Classification And Regression Tree). Therefore, macro level data from the municipalities of São Paulo state (Brazil) were used, such as demographic and socioeconomic data, fatality rates and other variables related to traffic. The results indicated the Gross Domestic Product (GDP) as the most important variable, and the municipalities were characterized mainly considering the size of the highway network and vehicle fleet (trucks, minibuses, cars, motorcycles). These characteristics could provide support to the government to plan mitigating actions in municipalities with the highest tendency to high traffic fatality rates.