Research On Data Restoration And Prediction Of Expressway Lane-Level Traffic Volume Based On Machine Learning

The rapid economic development has led to an increase in people’s demand for transportation,resulting in an imbalance between the supply and demand of road facilities,frequent traffic congestion,traffic safety and other problems,making the road traffic environment increasingly deteriorating and hindering the development of the city.The proposal of ITS effectively improves the running condition of urban roads and improves the level of road service.Now that the transportation industry has entered the era of big data,in the face of massive traffic data,rational use of data resources,in-depth mining of information,and real-time and accurate prediction of road traffic flow can improve the ability to induce and control urban traffic.Based on the multi-dimensional characteristics of expressway traffic flow in the spatiotemporal distribution,this paper constructs a data restoration model based on an ensemble learning algorithm and a prediction model based on a deep learning combination algorithm,which provides a theoretical basis for the management and control of expressway traffic.Firstly,the domestic and foreign researches on the restoration of missing traffic flow data and the short-term prediction of traffic flow parameters are sorted out,and the existing research methods are summarized and evaluated,and it is concluded that most of the current research objects in restoration and prediction are road section traffic flow.In order to meet the increasingly refined needs of traffic forecasting,this paper studies the traffic flow in the lane,and determines the main research content and technical route of this paper.Secondly,the data source,data content and preprocessing method of the traffic flow microwave data used in this paper are briefly introduced,and the traffic flow characteristics are analyzed based on the experimental data.On the one hand,analyze the basic characteristics of the three parameters of traffic flow;Then,the missing of traffic flow data is divided into three categories: missing point,missing line and missing area.Then,for the two types of missing points and missing lines in lane-level traffic flow data,a repair model based on ensemble learning algorithm is proposed.The XGBoost algorithm,which has good performance and can handle missing values in the ensemble learning algorithm,is selected as the basic repair model,and the IGA algorithm is introduced to optimize its parameters,and the IGA-XGBoost repair model is obtained.The repair effect of the model under two missing types is tested,and a new solution is provided for the repair of missing traffic flow data.Finally,in order to fully exploit the distribution characteristics of lane traffic flow in the spatiotemporal range,this paper constructs a deep learning combined model based on graph convolutional neural network and gated recurrent unit,namely the GCN-GRU model.The model uses a graph convolutional network to capture the spatial distribution characteristics of traffic flow,and uses a gated recurrent unit to obtain the temporal distribution characteristics of traffic flow.The experimental verification is carried out with the lane traffic flow data on the expressway section as the experimental data.The results show that the GCN-GRU model has a better fitting effect,and the prediction accuracy is higher than that of the benchmark models such as GCN and GRU.There are 43 pictures,18 tables and 84 references.