Research On Prediction Of Truck Travel Time For Expressway Based On Random Forest Model Of Particle Swarm Optimization

National development should prioritise transportation first.With the continuous improvement of the expressway network structure and the rapid development of the logistics transportation industry,the total quantity of cargo vehicles has been increasing,continued to add pressure on the operation and management of expressway,expressway congestion,traffic accidents and other conditions,reducing the speed and quality of goods transport.In light of this,China will strengthen the research and construction of intelligent transportation systems,improve the level of the intelligent transportation system,and use high-tech information technology to ease traffic pressure.Expressway freight transportation is an indispensable part of transportation modernisation.Combining it with intelligent transportation can improve transportation efficiency,reduce logistics costs,and promote the development of the logistics industry.Travel time prediction is the basis of intelligent transportation systems.The travel time prediction of the expressway enables travellers to know the traffic situation in advance,ensures cargo carriers to plan the travel path,reduces the transportation time cost,and improves logistics transportation efficiency.By controlling the changing trend of traffic,expressway users and managers can make better use of expressway resources,rationally optimise the allocation of resources,promote the modernisation and informatisation of the transportation industry,and provide a broader space and better service for the development of expressways.At present,the research on travel time does not distinguish the vehicle type,and most of them are type one passenger cars.While the proportion of cargo transportation in the transportation economy is increasing,cargo transporters are more sensitive to travel time,and the difference in the attributes of trucks and vehicles has a great impact on travel time.This thesis focuses on the prediction of freight car travel time based on the toll collection data of the expressway network and the difference in the attributes of different types of vehicles.Firstly,the basic toll data of the expressway is analysed and preprocessed,and the method of abnormal data identification and processing is proposed in this thesis.Besides,special processing is carried out according to the truck type and other characteristic attributes so as to provide accurate and effective data for the subsequent travel time prediction.Through the preprocessing of expressway toll data,the characteristics of expressway traffic flow are deeply analysed,and the related factors affecting the prediction of expressways travel time,such as traffic volume,vehicle type and historical travel time,are explored.Then,regardless of the influence of vehicle type,all vehicles were converted into standard vehicle equivalent numbers for mixed flow traffic characteristics,and a Random Forest(RF)prediction model was adopted to predict the average travel time of vehicles on expressways.The influential factors related to expressways travel time,such as traffic flow and historical travel time,were analysed.Pearson correlation coefficient and grey correlation analysis were used to determine the influencing factors of travel time.The traffic flow and historical travel time were mainly used as the input feature vectors of the model to construct the prediction model of random forest travel time based on mixed traffic flow.A particle swarm optimisation algorithm was introduced to optimise the stochastic forest model,and the optimal combination of parameters of the stochastic forest was determined.In this way,the Stochastic Forest Model(PSO-RF)based on the particle swarm optimisation algorithm was constructed to predict the travel time of the expressway.Taking Jiuyong Expressway S7 Gaoxinnan toll station to Yongchuandong toll station as an example,the predicted result of the particle swarm optimisation model is reduced by 0.711% compared with the unoptimised average percentage error(MAPE).At the same time,the prediction results of the historical average model,BP neural network model and SVR model were compared and analysed.The MAPE of the RF model was reduced by 2.137%,1.282%,and 4.738%,respectively.Compared with the comparison model,the MAPE of the PSO-RF model is reduced by 2.848%,1.993%,and 5.449%,respectively,which verifies that the RF model and PSO-RF model proposed in this thesis have good effect in the prediction accuracy of expressway travel time.Finally,considering that the travel time of trucks is affected by the characteristics of vehicles and the volume of external traffic,the characteristics of trucks on expressways are not only different from that of buses but also greatly different between different types of trucks.Through analysis,it is found that the travel time increases with the increase of the proportion of truck types and the load weight of trucks.The travel time characteristics of trucks on expressways were analysed.One-hot coding was used to highlight the differences between passenger and cargo and the characteristics of vehicles.The model was added as the feature vector,and the Random Forest Model(PSO-RF)based on the particle swarm optimisation algorithm was constructed to predict the travel time of trucks on expressways by vehicle types.Taking the Jiuyong Expressway S7 Gaoxinnan toll station to the Yongchuandong toll station as an example,the prediction results of the historical average model,BP neural network model and Support Vector Machine Regression Model(SVR)were compared and analysed.No matter what kind of truck type,the performance of the multi-feature PSO-RF model is better than that of the comparison model in the three evaluation indexes of MAE,MAPE and RMSE,which proves that the PSO-RF model has good effectiveness in predicting the travel time of different types of the truck on the expressway.In conclusion,this thesis verifies the feasibility and accuracy of the random forest model based on the particle swarm optimisation algorithm in the prediction of expressway travel time and opens up a new angle of expressway truck travel time prediction.The study of this thesis can provide valuable traffic information for expressway travellers and freight carriers and plan travel routes according to the pre-predicted travel time to avoid congestion and save travel costs.It is helpful for traffic management departments to effectively control the changing trend of expressway traffic flow and take timely guidance and control measures.It can also provide references and ideas for existing expressway travel time prediction methods and has good practical application value and social benefits.