Aplicação de redes neurais artificiais na previsão de acidentes rodoviários em pistas simples: estudo do trecho da BR-230 pertencente ao estado da Paraíba.

Abstract

In Brazil, road accidents are one of the major social and public health problems in the country. In Paraíba, in particular, it is no different, the path that connects the city of Campina Grande to the state's backlands brings expressive numbers. One of the forms of mitigation that has been studied is the application of neural networks for the prediction of accidents, the objective of this project being to develop an artificial neural network model to predict the frequency of accidents on single-lane highways on the BR-230 with stretch from the municipality of Campina Grande to the municipality of Cachoeira dos Índios, in the state of Paraíba. For comparative purposes, three different forms of data analysis were used from a neural network chosen based on the results obtained in tests, which correspond to the analysis with error margins with ranges ranging from 5 to 30 km, the analysis dividing the highway in stretches of 5 to 30 km and the third analysis consisted of dividing the highway into stretches between municipalities. This analysis procedure by different methods allowed us to conclude that the best way to assess stretches where intervention with safety measures is necessary is to divide the highway between its municipalities, resulting in larger extensions with a higher rate of correctness for the neural network. Based on the results it was possible to observe that the stretch where accidents occur most is between the municipality of Sousa and Cachoeira dos Índios, passing through Cajazeiras, where there is a greater concentration of cases taking into account the proportionality of the stretch.