Prediction And Applied Research Of Short-term Traffic Flow Based On Neural Network Model

In recent years, China has entered a period of rapid development ofurbanization. The increasing number of motor vehicles has brought greatconvenience to people’s life, but also brought the traffic congestion,environmental pollution, resources waste and other issues at the same time. Thiscaused serious impact on people’s daily life and attracted the attention of manyscholars. Intelligent transportation system is a solution to ease traffic congestion,reduce resource consumption and environmental pollution through theinteraction of people and information system. Internet of Vehicles is a typicalapplication of Internet of Things technology in the field of intelligenttransportation. As one of the key technologies, the main function of the trafficflow prediction is to provide traffic guidance and assist to reduce the travel timeand traffic congestion based on road traffic information.Short-term traffic flow prediction is one of the key research of traffic flowprediction, timely and accurate short-term traffic flow prediction information is guarantee to smooth traffic. Short-term traffic flow information is self-similar, ie,traffic information from different sections of the same period show a certainregularity and periodicity, which provided favorable conditions for short-termtraffic flow prediction; at the same time, short-term traffic flow informationpossess the characteristics of real-time, high-dimensional, nonlinear,non-stationary. However, the existing traffic flow prediction algorithms tend tohave low accuracy, slow convergence, unstable performance, the problem ofshort-term traffic flow prediction was studied in this thesis, and correspondingalgorithms were put forward. The main contributions of this thesis include thefollowing aspects:Firstly, This thesis analyzes the dynamic traffic flow guidance model basedon the study of existing traffic flow prediction methods, and collects traffic flowdata by the traffic simulation software VISSIM, and gives the curves of trafficsimulation input&acquisition and average speed parameter.Secondly, this thesis analyzes the basic characteristic parameters of trafficflow by combining methods of historical trends and the adjacent padded;repaired the wrong data and reduced the impact of erroneous data on theprediction accuracy of the model by traffic flow data preprocessing.Finally, a new short-term traffic flow prediction algorithm GA-BBP forwas proposed in this thesis. GA-BBP algorithm selects BP neural network as thebasis for short-term traffic flow prediction model, methods of dynamicadjustment of learning rate and Bayesian training are adopted, and an improved Genetic Algorithm is introduced to optimized the parameters. GA-BBPalgorithm has higher accuracy and convergence rate than traditional algorithm,simulation results show GA-BBP algorithm is more accurate and robust.