APRENDIZADO DE MÁQUINA UTILIZANDO AGRUPAMENTO E REGRESSÃO NA PREVISÃO DE LOCAIS DE ACIDENTES DE TRÂNSITO EM ZONAS URBANAS

Abstract

With the urbanization of Brazilian cities, automobile locomotion has become indispensable, so the area of urban mobility has increased on an exponential scale, resulting in an increase in traffic violence, whether caused by traffic jams, human bias or infrastructure problems. This work proposes a solution that predicts accident locations within urban areas based on temporal data (date and time) of accidents. It uses the K-Means algorithm to group and KNN Regressor to predict, within the sample of accident data from the city of São Paulo collected between 2019 and 2021, a predictive model with an accuracy of 96.04% within a tolerance of 500m was obtained.