Study On Traffic Accident Forecast Method Based On Grey Theory Neural Network

Traffic safety is an important issue of common concern to all countries in the world, theroad traffic accident is a major influence to road traffic safety. Road accidents forecastingaims to estimate and predict the situation of road traffic in the future, through the past andpresent status of road traffic accident in the system, and considering the changes of its relatedfactors, then make the procedure to describe the future state of road traffic accidents. Roadtraffic accident forecast is of important practical significance for road traffic safety evaluation,planning and decision-making.Road traffic system is a complex system which is affected by many factors, at the sametime, it has less samples, poor information and non-linear features. The traditional linearprediction method has large limitation in solving nonlinear problems, therefore, the trafficaccident forecasting accuracy is not high. Considering the advantages of gray system model tosolve the problem of "small sample, poor information" and the neural network model canapproximate any nonlinear function, this paper applied the two forecast methods to establishthe grey GM (1,1) forecasting model and neural network forecasting model.Traffic accident forecast need to select influential factors for the traffic accident which isknown as predictors, and determine the target of prediction which is called the forecastingpredictors. This paper uses principal component analysis to calculate the correlation indexbetween traffic accident influence factors and the comprehensive traffic accident mortality,and select the factor which has big correlation with comprehensive traffic accident mortalityas the impact indicators. And then according to the correlation analysis with impact indicators,to determine the forecast target which is namely the forecasting predictor. Using these twosingle methods to do macro forecast traffic accident prediction in our country, the predictedresults verify the effectiveness of the two methods in the traffic accident prediction.The comparative analysis of forecast result shows that single forecasting modelprediction accuracy is not high. Finally, we combine the advantages of two models to buildthree gray neural network combination forecast models. Through the comparative analysisresults, we arrived at the conclusion that combination forecast results traffic accidentprediction accuracy of grey neural network method is better than single prediction methods.