ABSTRACT
To predict the development of road traffic accident scientifically and accurately, traffic fatalities number of death in China from 2002 to 2015 are taken as a sample. The optimal smoothing coefficient and the optimal initial smoothing value are determined by nonlinear programming method and curve fitting method respectively. Then optimal cubic smoothing model is established to predict the traffic fatalities number of death from 2016 to 2018. Taking −2%, 4%, 10.5% and 17% as the state partition thresholds, and the Markov model to revise the prediction results. Finally, the results are compared with Markov's modified traditional exponential smoothing model. Results show that the prediction accuracy has improved by 0.2%, with the average relative error decreases from −5.01% of the traditional cubic exponential smoothing model to −4.81% of the optimal cubic exponential smoothing model. And the prediction accuracy has even improved by 1.16%, with the average relative error decreases from −3.27% of the Markov's modified traditional cubic exponential smoothing model to −2.11% of the proposed model. The proposed model presents superiority and can be used to provide scientific and accurate reference for traffic safety management in reality.
