Mineração de Dados na Construção de Modelo de Predição de Acidentes com Vítimas em Recife

Abstract

One of the prominent urban questions in Recife is traffic. Here we intend to aid mitigating this issue, predicting accidents with victims. To achieve this goal, data from CTTU, available in Portal de Dados Abertos da Prefeitura do Recife were used. These data cover the accidents that happened between June 2015 and February 2020. So, this information went through a data mining process. Hence, a machine learning prediction model for traffic accidents with victims in last 12 months in Recife was created. Combating the COVID-19 caused movement restrictions that changed the expected traffic profile in the city starting in March 2020, reason for the exclusion of the period in this model. We proposed four prediction models and, for the model with the best accuracy, the average error rate was 13 accidents per month.