Accident Cause And Prediction Model Research Of Road Traffic Safety

As an important means of study on road traffic safety, prediction method research of road traffic safety has attracted more and more concern and attention. By analyzing the existing data, suitable quantitative models have been constructed which can reflect system development regularity to achieve the goal of mastering the future development trend of road traffic safety, so as to take corresponding measures timely to avoid the potential risk as well as providing the reliable basis for making the road traffic safety goals keeping pace with the times reasonably.Road traffic system has many characteristics,such as dynamic, randomness, causality and reproducibility, which make improving road traffic safety more complicated.Starting with the china and abroad mature genetic theory of road traffic accident, causes of road traffic accident have been analyzed on four aspects: man, vehicle, road and environment, the relationship between the road traffic safety and the main factors which influenced the road traffic safety has been discussed, so this thesis has provided a theoretical basis for strengthening management of road traffic safety. Prediction method of road safety based on the grey theory has been studied on by analyzing the existing safety prediction models and methods.(1)Prediction of grey theory has many remarkable advantages ,such as less sample, easy calculation, high accuracy and strong practicality etc, but it can't do good performances when complexity of sample data enhanced with the increased of them. Because the grey GM(1,1)model shows exponential growth, which mainly suitable for the sequences that according to single exponential increase law. So, improved GM(1,1) model based on the equal dimension and new information has been proposed when excessive sample data problem appeared.(2)By using the fusion and penetration ability of grey prediction model, combination has been made with the other single model to get the combination forecasting model, in order to realize mutual compensation of information and advantages. Thus, not only inescapable limitation of single model can be avoided, but also the prediction accuracy can be guaranteed effectively when the superiority of each single model has been integrated, which makes the prediction more feasible and reliable.(3)Grey neural network combination forecasting model has been built, which based on the grey equal dimension new information model and BP neural network. 1982~1998 national accident number predictive value has been analyzed as an example by using least square method to determine weight coefficient of each single model in the combined model. The results show that: the precision of combined model is superior to the single model.(4) GM(1,1)model can't have good enough performance in prediction precision when the sample data is extraordinarily stochastic and fluctuant. According to this weakness, markov chain theory has been introduced based on the grey prediction theory. Grey markov prediction model according to the 1996~2005 death toll of road traffic accident has been built to confirm the practicability and high effectiveness of the grey extended model based on markov chain theory.Take the road traffic safety management as a starting point in the thesis, mature genetic theory of road traffic accident as well as the typical single prediction model has been analyzed comparatively. By analyzing the examples of some road traffic safety index, the feasibility and reliability of the prediction model which is based on the grey system theory has been verified when it is used in road traffic safety prediction.