SEASONAL MODELLING OF ROAD TRAFFIC ACCIDENT

Abstract

This work modelled road traffic accidents that exhibit seasonal behaviour and non-statioarity in variance. The modelling procedure was demonstrated using monthly road traffic accident data (1999-2014) obtained from federal road safety commission in Port Harcourt, Nigeria. Seasonality and non-stationarity in variance were detected from the series raw plot, autocorrelation and partial autocorrelation functions. To obtain stability in variance and level of the series, some transformation techniques were applied. Seasonal component was incorporated into the Box and Jenkins (1970) ARIMA model to cater for the periodic nature of the series. The resulting estimated seasonal-ARIMA model was subjected to different diagnostic checks, and was found to be adequate. The proposed model was then used to generate forecasts of road traffic accidents for the next thirty months.

Keywords: Stationarity, Seasonality, Transformation, Autocorrelation and Partial autocorrelation.