Forecasting the Fatality Rate of Traffic Accidents in Jordan: Applications of Time-Series, Curve Estimation, and Multiple Linear Regression Models

This paper investigates the trends in the traffic fatality rate per 100,000 population, population growth, gross domestic product (GDP), and registered vehicles per 100,000 population over 39 years (from 1981 to 2020). Traffic accidents data were obtained from the Jordan Public Security Directorate (JPSD) published reports for the selected years in Jordan. Data were analyzed to predict the annual fatality rate using time-series, curve estimation, and multiple linear regression models. Among the various available models for curve estimation, the cubic model outperformed the rest by capturing 79.4% of the variance. Also, multiple linear regression results showed that increasing the length of the road network can play a role in decreasing the fatality rate of road accidents. While time series analysis offers numerous techniques, it is determined that the Jordanian fatality rate is best suited for using the exponential smoothing approach. Results indicated that the time series model produces the lowest mean absolute percentage errors (MAPE), followed by multiple linear regression, and finally by the curve estimate (cubic) model. It is essential to see how these variables have changed over the study period, which helps decision-makers, engineers, and researchers predict future trends and suggest suitable measures to lower the fatality rate. © 2022 School of Science, IHU. All Rights Reserved.