Regras de associações entre as características dos acidentes de trânsito em rodovias federais brasileiras por meio de aprendizado de máquina

Resumen

Traffic accidents are considered a serious public health problem and the significant number of deaths highlights the need for a deeper analysis of the causes of accidents. The objective of this research was to identify rules of association between the causes of accidents and the characteristics of vehicles, roads, users and the environment on Brazilian federal highways. The machine learning techniques Apriori, Eclat, FP-Growth and FP-Max were compared. The methodology proposes a data table of categorical variables, in a mixed method for data collection and transformation. A case study was carried out within a real context. The comparison between algorithms and conclude that Apriori, FP-Growth and Eclat present the same performance, with similar support indexes and amount of characteristics. The FP-Max in reverse, providing a more accurate result. The study presents association rules such as, for example, a male driver, driving who drives a vehicle on a non-holiday day, outside traffic hours, on a straight line, is associated with accidents where the cause is not keeping safety distance.