Travel Time Prediction And Its Application Based On Guangxi Highway Toll Data

With the rapid development of social economy in China,the number of cars keeps increasing,the demand for public cross regional travel is increasing,the demand for highway is far beyond its construction speed,which makes the highway bear more and more traffic pressure,which leads to frequent road congestion and traffic accidents,and the overall service level of highway is greatly reduced.In order to alleviate the pressure of transportation brought by economic development and optimize resources,China has increased the research and construction of its.The prediction of the travel time of highway is one of the important contents of the construction of intelligent transportation system.Based on the data of toll collection of Guangxi highway,this paper proposes a travel time prediction method suitable for Guangxi highway.The main work completed includes the following three points:(1)A data cleaning method based on the characteristics of Guangxi highway toll data is proposed to improve the quality of the data set.Firstly,using the exit time and entrance time fields in the original export charge data table to calculate the travel time of each vehicle.Then eliminating abnormal data according to a certain rules and filtering valid data by the "3σ" principle of normal distribution.After that,simplify the data table,only leaving the fields needed for the research.(2)A travel time prediction method suitable for Guangxi highway is proposed.Firstly,the feature engineering of travel time is constructed.Taking the average travel time of the first three periods of the prediction period,vehicle type,vehicle classification,week and hour as the feature variables,the maximum and minimum standardization method is used to normalize the numerical features,and one-hot coding is used to process the classified features;Then,the random forest algorithm and BP neural network are used to establish the travel time prediction model of highway.Finally,the flow data from Liujing to Nanningdong of Guangxi highway is taken as the research object to train the model,verify and evaluate the prediction results.The result shows that the BP neural network is more suitable for building a highway travel time prediction model than the random forest algorithm.BP neural network has better generalization ability.The average absolute percentage error(MAPE)and Mean Absolute Error(MAE)are 6.23% and 116 seconds respectively,which are within the acceptable error ranges.(3)The travel time prediction model is applied to the "intelligent highway integrated platform" system.The travel time prediction model is built based on Python language and python framework.The cross language and cross platform communication between the "intelligent high-speed integrated platform" system background and the prediction module is realized through socket.The prediction function is achieved by using Spring to call the service provided by the python process.The research results of this paper are applied to the actual production,so that the management department can understand the trend of highway traffic change in time,and do a good job in traffic guidance and control;The travelers can plan the driving route in advance according to the predicted time,effectively avoid the congested road sections,save the travel cost,reduce the energy consumption of vehicles,reduce the exhaust emissions and noise pollution,which has good social benefits.