Jordan University of Science and Technology

Forecasting Traffic Accidents in Jordan Using Regression Techniques

Authors:  Mohammad Ali Khasawneh, Aslam Ali Al-Omari, and Bader Ganam

The data used in this study covered the period of 1981-2015. The data represented yearly accidents, injuries, fatalities, registered vehicles and population. The analysis of accident forecasting was concerned with building regression models for the prediction of accidents until 2030. In brief, the regression models could predict the yearly numbers of registered vehicles, accidents, injuries, fatalities and financial cost of accidents according to five models. Simple linear regression model, curve regression model and Smeed?s equation model were performed in the analysis. All regression models were good with high coefficients of determination (R2) that ranged from 99.3% as the highest value for registered vehicles? model to 79.1% as the lowest value for fatality rate model with significant parameters under a confidence level of 95%. Regarding forecasting these models for the period 2016-2030, the prediction rates of accidents in 2030 exhibited good improvement in general when compared with the rates of 2015. The rate of injuries per 1,000 accidents and number of fatalities per 1,000 accidents are predicted to go down by about 77 injuries and 2 fatalities from 2015 to 2030, which indicates good behavior in traffic safety measures.