Prediction Model Of Short-term Traffic Volume Based On Spatially And Temporally Correlated State Of Multi-sections Of Section

With the rapid development of urbanization process, the motor vehicle is increasinglypopular, people enjoy motor vehicle of the enormous convenience while also facing thetraffic congestion confusion, traffic accidents brought about due to traffic congestion, airpollution, noise pollution,resource shortages and other problems, has increasingly becomethe main issues to hinder the global economy and even threaten human survival, has thusbecome the focus of widespread concern of governments around the world. For many cities,road congestion, bad traffic is in fact not just because the road is not enough, but the roadhas not been well utilized. Advanced traffic management and control systems for themitigation of urban road congestion is of great significance. Predict future road trafficstatus, based on current traffic conditions of the urban road network, traffic participantsactively cooperate to optimize driving routes to avoid the congested road, and give fullplay to the effective performance of the road network is of great significance for thebalanced distribution of traffic flow.Firstly, through an overview of the existing predictive models, comparative analysis,come to the advantages and disadvantages of various types of model forecast analysis.More commonly used methods and models are only for the time variable and a singleintersection section prediction analysis, not very well reflected in the temporal correlationbetween multiple sections of section in the road network of traffic flow. Kalman filter hasunique advantages: a wide range of adaptive Kalman filtering to a more flexible recursivestate space model can deal with smooth data, and can also handle non-stationary data; aslong as the state variables for the different assumptions on allows it to describe and dealwith different types of problems; model with a linear, unbiased, minimum mean squareerror of; model is easy to implement on a computer, and greatly reduces the computerstorage and computation time, suitable for online analysis; prediction accuracy higher.Correlation analysis of multi-section of section of the road network within the scope of thestudy, the use of cluster analysis methods to analyze the spatial and temporalcharacteristics of the road network, traffic flow data to identify a set of road sections in ahigher correlation unified forecast analysis multidimensional time model; combination of multi-dimensional time series to establish the data, and transformed into state space model,the establishment of multi-sections of section short-term forecasting model based onspatially and temporally correlated state last use of the Kalman filter theory and solve theobtained multiple sections of the road network section short-term traffic forecasts.Finally, select the part of the urban road network in Xi’an verified as an example,traffic volume data obtained by the survey, the input to the Kalman filter equations derivedto predict the period of time traffic data, and error compared with the real dataanalysis,forecast error of the results within the acceptable range.The proposed sections of multi-section based on the spatial and temporal associationto establish short-term traffic forecasts, and select the Kalman filter theory to model themore accurate and objective sections of section traffic flow prediction, at the same timehave a good engineering practiceapplications.