| With the continuous development of society,the quality of life of people is getting higher and higher,and the number of private cars is increasing rapidly,which makes the problem of road traffic congestion increasingly prominent.Moreover,highway traffic safety is easily affected by the weather,leading to frequent traffic accidents.The emergence of Intelligent Transportation Systems provides ideas for solving traffic congestion.By establishing short-term traffic flow prediction model,it is possible to predict the changes in traffic flow and achieve intelligent transportation information.On the one hand,it provides convenience for the travelers,and on the other hand,it provides data support for traffic management departments to implement traffic control.Based on the data provided by the Pe MS website from a highway vehicle detection station in California.This paper analyzes the distribution characteristics of traffic flow,learns traffic flow prediction methods,and constructs a short-term traffic flow prediction model for the highway using Particle Swarm Optimization(PSO)algorithm to optimize Long Short-Term Memory(LSTM)considering the impact of weather factors.The specific research contents are as follows:Traffic flow processing and analysis.Firstly,obtain traffic flow data and weather data on the Pe MS website and Meso West website respectively,and set time intervals to construct time series.Secondly,analyze the temporal characteristics and correlation of traffic flow.Finally,the traffic flow data and weather data are preprocessed using preprocessing methods to improve the accuracy of the prediction model.The preprocessing process includes identifying and filling abnormal data,denoising data,and normalizing data.Based on this,establish a highway traffic flow prediction model.Build traffic flow prediction model.Firstly,aiming at the problem of determining the proportion of global search ability and local search ability of Particle Swarm Optimization algorithm,Rastrigin,Ackley and Griewank are selected as benchmarks to verify the search ability of standard Particle Swarm Optimization algorithm and compression factor Particle Swarm Optimization algorithm.The verification results show that the standard Particle Swarm Optimization algorithm has better search ability.Secondly,the correlation between traffic flow data and weather data is analyzed by correlation coefficient analysis method,and the weather factors that have great influence on traffic flow are obtained.Finally,the PSO-LSTM prediction model considering the influence of weather factors is constructed.The PSO algorithm is used to optimize the maximum number of iterations,learning rate and the number of hidden layer neurons of the LSTM network.Then,the LSTM network structure is built with the optimal parameter combination.80% of the data is input into the neural network structure for model training,and 20% of the data is used for model verification.According to the time characteristic analysis and correlation analysis of traffic flow data,it is necessary to predict the traffic flow on weekdays and non-weekdays respectively.The performance of the model can be verified by analyzing the prediction results.Model validation and analysis.The error function Mean Absolute Error(MAE),Root Mean Square Error(RMSE),and Mean Absolute Percentage Error(MAPE)are selected to evaluate the advantages and disadvantages of the model.In order to verify the feasibility and effectiveness of the PSO-LSTM model considering the influence of weather factors proposed in this article,the LSTM model,the PSO-LSTM model without considering the influence of weather factors,the BP neural network model,and the Support Vector Machine(SVM)model are compared with the model in this paper.The results show that the MAE,RMSE,and MAPE of the model proposed in this paper are the smallest among the five models for predicting traffic flow on working and non-working days,indicating that compared to other models,the prediction model constructed in this paper improves the prediction accuracy and has better prediction performance. |
