Machine Learning Based Prediction Of Traffic Status Of Main Road Network And Visualization Research

With the rapid development of economy,the process of urbanization is speeding up and the living standards of the people are continuously improving.The number of urban vehicles is increasing rapidly,and the urban transportation capacity has been unable to meet the increasing demand for traffic.The problem of urban traffic congestion has become particularly prominent.It not only prolongs people’s travel time,brings inconvenience to life,but also causes huge economic losses,and seriously affects social progress and economic development.Starting from the traffic history data,this paper studies the traffic condition prediction of urban road network.Visualization of predicted results and historical traffic conditions.According to the forecast results,people can choose the appropriate way to travel and travel routes,so not only can effectively avoid congestion,save travel time,but also help to alleviate traffic pressure,make full use of road resources.Based on the history data of floating vehicles in Beijing,this paper proposes a method based on machine learning combination classifier to predict the urban trunk road traffic condition.The method considers that the current road condition has a strong spatial and temporal correlation with the surrounding road condition.In the space,the road network topology is constructed to extract the relevant sections of the predicted range within the predicted time.Assuming the current time is the time t₀,the traffic state of the relevant section t₀-t at the time is selected as the main feature of the prediction section.By using the machine classification algorithm,a single prediction model is built for each predicted section,and then a combined classifier is constructed to realize the short-term prediction of the traffic status of the backbone network in Beijing.This paper also uses visualization technology to visualize data in a graphical way for intuitive analysis.The main research work and achievements of this paper are as follows:1)Based on the traffic data of Beijing City,the backbone network is extracted,and the topological link of Beijing road network is constructed.The reachable range of each section in the prediction time is established,which constitutes the relevant section of the road to be predicted.Through the relevant sections of the road traffic data at t₀-t,to predict the characteristics of the road to filter.2)Through the comparative analysis of support vector machines,Bayesiannetworks,decision trees and neural networks,a separate classifier is constructed for each predicted section,On this basis,the Adaboost classifier is used to form a strong classifier.And the short-term prediction model of traffic condition is constructed to further improve the accuracy of short-term prediction.3)With the help of visualization technology,the predicted results and historicaltraffic data are visualized.Using 2D and 3D visualization technology,construct an intuitive,three-dimensional,interactive visualization model,which greatly facilitates the analysis of multi-dimensional data,that is more conducive to a deeper relationship of data.In this paper,through establishing the topological connection of the road network,the surrounding traffic data related to the short-term traffic condition of the road to be predicted is determined.The characteristics of the prediction section are screened and reduced.The machine learning method is used to build a prediction model for each section,and the short-term prediction of traffic status is realized.Besides,the visualization method is used to demonstrate,analyze and evaluate the traffic condition prediction effect of backbone network.