Aplicação de métodos de análise de séries temporais em dados de acidentes aéreos

Abstract

This study focuses on the application of time series analysis to data related to the monthly number of air accidents that occurred in Brazil from January 2012 to March 2023, aiming to predict the quantity of accidents and to identify a possible trend in the data. The study addresses both the general objective of fitting a time series model and the specific objectives of selecting the most appropriate model, applying the Box and Jenkins methodology, and performing a 12-step forecast. For this approach, the iterative Box and Jenkins cycle was used for each stage of model creation: model identification, parameter assessment, diagnosis of selected models, and finally, forecasting. The results suggest that the number of air accidents remained relatively constant over the analyzed period, with the identification of a SARIMA(2,0,0)(1,0,0) model as the most adequate to represent this series.