A Comparison Between Seasonal Autoregressive Integrated Moving Average (SARIMA) and Exponential Smoothing (ES) Based on Time Series Model for Forecasting Road Accidents

Abstract

Road safety guidelines are not properly implemented and are not diverse enough to counter an annual increase in traffic volume. The mitigation techniques of road regulating bodies have failed in minimizing road accidents with the increase in road users. Therefore, the purpose of this study is to provide valuable insight to the facilitators and decision-making stakeholders by predicting the number of accidents because it is an existential hurdle toward the prevention of accidents. Therefore, this study aims to create temporal patterns to forecast the accident rates in Pakistan by utilizing univariate time series analysis such as seasonal autoregressive integrated moving average (SARIMA) and exponential smoothing (ES) models. The results indicate that the ES model fitted better on accident data over the SARIMA model after calculating the lowest mean absolute error, root mean square error, mean absolute percentage error and normalized Bayesian information criterion. The study provides the guiding principles to implement the forecasted accident rates in the designing of roads to ensure the safety of end users which is a prime interest for accident rate collection agencies, decision-makers, design consultants and accident prevention departments.