Research On Traffic Volume Prediction Based On Highway Toll Data

The rapid growth of highway travel demand leads to highway traffic congestion and traffic safety problems.In order to alleviate traffic congestion and improve traffic safety level,it is necessary to strengthen the intelligent control of highway traffic.The accuracy and reliability of traffic flow prediction directly affect the formulation and implementation of traffic control strategies.By processing the toll data within the range of the expressway network in a province,the thesis analyzes the temporal and spatial rules of expressway traffic flow within the road network area,studies the short-term prediction method,and provides support for the formulation of intelligent expressway control strategy.The main research work of this thesis is as follows:(1)Describe and analyze the original highway toll data,extract data related to traffic flow status,and generate highway travel data set(data set for short).The data set is processed,the highway traffic characteristics are mined and analyzed,and the factors affecting the highway OD traffic volume are determined.The highway toll station is taken as the virtual traffic cell,and the OD matrix within the range of the regional highway network is generated.In view of the advantages of LSTM in processing time series data,a LSTM-based OD traffic volume prediction model is constructed,and OD traffic volume prediction under different time granularity is studied.(2)Considering the large difference in traffic volume between OD pairs,in order to fully learn the change characteristics of different traffic,K-Means clustering algorithm is used to cluster OD pairs.According to the clustering results,LSTM models are constructed for different categories of OD respectively for training and prediction,and good results are obtained.(3)In order to further verify the effectiveness of Kmeans-LSTM model,a regional road network is selected to build a traffic flow distribution model,and the predicted OD traffic volume is allocated to each section of the road network.By comparing the predicted traffic volume with the actual traffic volume,the results show that the error is small,which verifies the validity of the forecast model and the distribution model.