911. Qual é sua emergência?: modelagem de séries temporais para previsão de demandas de ligações para centrais de emergências

Abstract

This study aimed to forecast the daily number of emergency service calls in Baltimore. To achieve this goal, the methodology proposed by Box-Jenkins was employed, and data collected from 2015 to 2021 were used. Initially, calls received at the 911 telephone number were classified according to their nature, encompassing categories such as health emergencies, traffic incidents, and fires. However, for the analysis, these categories were consolidated. Following a detailed exploratory analysis of the data, two SARIMA models were developed. These models incorporated a daily seasonal variability of 7 days, with parameters that proved to be significantly different from zero. Furthermore, it was verified that these models were in the appropriate stationarity and invertibility region for accurate forecasting. The next step was to ensure that the residuals' vector of the models behaved like a white noise process, ensuring the reliability of the generated forecasts. Subsequently, forecasts were estimated based on the developed models. By evaluating forecast error statistics, it was possible to identify and select the model that proved most suitable for forecasting purposes. This systematic and rigorous analysis process allowed for obtaining reliable results in forecasting the daily number of calls for the emergency service in Baltimore