Research On Data Mining And Prediction Method Of Traffic Flow

The real-time,accurate and reliable prediction of traffic flow is the key to realize the control of the intelligent transportation system.City road traffic flow is vulnerable to the surrounding factors,it has high noise,nonlinear,complex network characteristics.Traditionaltraffic flow forecastingmethods,including support vector machine,wavelet theory,neural network and time series analysis,are generally derived from other research fields.Using these prediction methods to predict traffic flow directly,but not considering the traffic flow data characteristics,often leads to poor prediction results or different prediction accuracy due to different roads.Starting from the theory of grey system and grey modeling mechanism,we will expand the data ratio than for the generalized data ratio.According to the nature of generalized data ratio,a new extreme point judgement method is proposed in this paper,which is used for data mining of traffic flow and a new adaptive step grey model for traffic flow prediction.The specific work is as follows:1.The definition of generalized data ratio is proposed.Grey sequence generation is one of the important research methods of grey system theory,grey modeling is the basic law of accumulated generating grey model,ratio is an important indicator to determine whether the data sequence exponential transform.Based on the concept of data ratio in grey theory,the definition of generalized data ratio is proposed in this paper to describe the direction and rate of change of time series data more effectively.2.A new method of judging extreme value points is proposed and used for data mining of traffic flow.By comparing time series representation,a method based on generalized data level ratio to estimate extreme points of time series is proposed.Compared with the traditional extreme point judgment methods,this method has high accuracy in judging extreme points.This method is used to segment the traffic flow of the US Minnesota,and then the descriptive statistics,distance measurement and classification recognition of traffic flow segmented data are carried out.3.A grey model prediction method with adaptive step length is proposed,and it is used to predict the short time traffic flow.The data characteristics of city traffic flow based on the daily cycle of change,and the approximate relationship between generalized data ratio and the parameters of grey model,put forward a kind of automatic selection step gray model algorithm.In this paper,two traffic flows in the United States Minnesota are used to compare the performances of the different models: the modified grey model,the traditional grey model,the historical average method and the moving average method.The simulation results show that modified grey modelcan significantly reduce the absolute error,and has higherperformances on forecasting.It is suitable for real-time forecasting of short time traffic flow in urban road.