Forecasting Traffic Accidents According to (Types of Roads and their Causatives) in Iraq using ARMA of Low Ordered Combinations Models

Abstract

Objective: This research aimed to forecasting of an expected outcomes based on the numbers of traffic accidents concerning (Types of Roads, and Their Causatives), in Iraq through applying ARMA of low ordered combinations models.

Methodology: Data are selected from annuls statistical reports traffic accidents registered and published by ministry of planning for the time periods (2002-2015). Two statistical criteria had been applied, such that (MSE, and MPE), first statistic measuring the magnitude of the errors, a better model will give a smaller value. The second statistic measuring the bias, a better model will give a value close to zero. As well as 95% confidence interval limits of predicating long term trend are estimated.

Results and Findings: The best reference of predicating traffic accidents according to types of roads, and having significant parameters estimates in at least at P<0.05, and as follows: For high ways, results shows that ARMA(0,1) is the best model, and the projection are stationary according to general mean line, then followed with main roads, which proved that ARMA(1,2) is the best model, and projection are decreased, then followed with subsidiary roads, which proved that ARMA(1,0) is the best model, and projection are increased, then finally Arcadian roads, proved that ARMA(1,0) is the best model, and projection are increased.

As well as, best reference of predicating traffic accidents according to Causative sand having significant parameters estimates in at least at P<0.05, and as follows: For road, results shows that ARMA(0,1) is the best model, and future projection are increased, then followed with cars, which proved that ARMA(0,2) is the best model, and the projection are stationary at the general mean line, then followed with drivers, which proved that ARMA(0,2) is the best model, and the projection are stationary according to general mean line, then followed with infantry, which proved that ARMA(0,2) is the best model, and the projection are stationary according to general mean line, then followed with cars, which proved that ARMA(0,2) is the best model, passengers, proved that ARMA(2,1) is the best model, and the projection are non stationary, then finally other causatives, proved that ARMA(1,0) is the best model and projection are decreased.

Conclusions: Among several forecasted low ordered ARMA models, unique combination are nominated for each status of studied traffic accidents, either for types of roads, or for their causatives, and that were accordance of studied statistical indicators, which assigned a significant level for estimated parameter's models, and make high reliability for predicates of accidents numbers in the future projections.