Research On Urban Road Intersections-accident Prediction Model And Algorithm

The road accidents forecasting is the important content of the research of traffic safety. It is divided into macroscopic forecast and microcosmic forecast.The microcosmic forecast includes the prediction of section and intersection accidents mainly,it has a more important significance.Once the unusual accidents presented in a certain intersection is predicted,the accidents can be reduced through safety measure,besides the effect after improvement also can be predicted,this will play a positive auxiliary part in reducing traffic accidents effectively,this will offer reference basis for making the precautionary measures of pertinence and traffic law.In this thesis, Firstly, the cause of traffic accident is analyzed form three aspects of people,vehicle and road environment.Then discuss the traffic accident happened mechanism according to accident catastrophe theory.Secondly,The existing intersections-accident prediction method is carried on the overall summary,both pluses and minuses with scope of application about various kinds of methods is defined,on this basis,the intersections-accident prediction method based on neural network is put forward.Research on the methods that how to determine the number of neural network neurons.Establish intersections-accident prediction model on the basis of neural network. Finally,the instance analysis is carried out with traffic accident census of shijiazhuang, the influence factor of intersection accident is selected as import variable,then the algorithm of neural network is actualized through Matlab progamme, the model is proved to be a effective and accurate intersection-accident prediection model.