Forecasting Traffic Accident Fatalities in the Kurdistan Region of Iraq using VARMAX Model

Abstract

 Traffic accident fatalities are one of the major global concerns, due to multiple factors affecting them and their impact socially and economically. Thus, forecasting the number of fatalities in any region is very significant in reducing this issue, especially when the number of vehicles is notably high and many families possess more than one car. Kurdistan Region of Iraq (KRI) has high fatality rates, as well as a high level of vehicle ownership in each governorate, highlighting the critical need for improved traffic management and safety strategies. This article aims to forecast traffic accident fatalities in KRI using multivariate dynamic time series models, specifically the Vector Autoregressive Moving Average with Exogenous variables (VARMAX) model. All the statistical analyses were performed using the Python programming language based on monthly data from 2015 to 2024. The average temperature and driving speed were included as exogenous factors to provide an extensive perspective of the dynamic regression relationships within these models, while the number of fatalities from three governorates (Erbil, Sulaymaniyah, and Duhok) were treated as endogenous variables. The dataset was divided into two subsets, with 80% used for training the model and 20% reserved for testing the performance. The results revealed that the VARMAX(1,1) model delivers the highest performance among the parameters investigated. Forecasts for traffic accident fatalities in the three governorates reveal a rising trend in 2025 and 2026. These findings underline the urgent need for targeted road safety programs and policy reforms to mitigate the prognosed increase in fatalities.