Abstract

In this work, time series models are applied to explain and forecast the rate of traffic accidents, deaths and injureds in Luanda, Angola. Monthly Luanda data from 2002 to 2015 are used to fit models and to make predictions. Road accidents in Angola are currently one of the major causes of death in the country. Particularly Luanda, the capital, is the province that shows the highest rate in terms of accidents, deaths and injureds. However, in recent years there has been a decrease in the accidents rate, with average growth rates of -6.73%, 0.19% and -2.54\% for accidents, deaths and injureds respectively. We have used classic Seasonal ARIMA models (SARIMA) in two different approaches, the first one treat all observations the same way. The second approach identifies outliers, taking into account its magnitude and estimates SARIMA models for the series excluding the significant outliers. A Seasonal-Trend decomposition based on a locally-weighted regression smoothing (Loess) approach was also applied. The SARIMA models that take into account the extreme values revealed to fit and predict better than the pure SARIMA models time series of traffic accident data.