Research On Short-term Passenger Flow Forecast And Empirical Analysis Of Urban Rail Transit

As our country’s urbanization process continues to accelerate and the urban population continues to increase,the problems of urban traffic congestion and environmental pollution are also becoming more and more serious.Urban rail transit has become the preferred means of transportation for urban residents due to its large capacity,low energy consumption,high punctuality rate,and environmental protection.In recent years,due to the accelerating pace of urban rail transit construction;a total of 40 cities in my country have opened 6,730.27 kilometers of urban rail transit operating lines as of December 31,2019,whose passenger traffic exceeds 24 billion passengers and hit a record high.The continuous increase in passenger traffic has brought a great test to the operating organization department.Therefore,short-term passenger traffic forecast has become a key issue that needs to be solved urgently in order to improve operational efficiency and ensure driving safety.The short-term passenger flow of urban rail transit has the characteristics of randomness,periodicity and correlation,and short-term passenger flow forecasting is a key technology for the optimization of urban rail transit operation organization.The accuracy of its prediction results will directly affect the urban rail transit operation organization rationality.Therefore,based on the historical passenger flow data of Hangzhou Metro,this article predicts the short-term passenger flow of urban rail transit.First,based on experimental data,in this paper,the statistical characteristics of urban rail transit passenger flow distribution are analyzed from the time and space dimensions,and the sample data is preprocessed.Then,taking a period of inbound passenger flow as an example,the Pearson correlation coefficient and K-Means clustering algorithm are used for correlation statistical analysis and classification,which makes an important foundation for subsequent research.Finally,we build a passenger flow prediction model based on XGBoost,LightGBM and LSTM.Using MAE,RMSE and MAPE as the performance evaluation indexes of the model,the parameters of each model are tuned through methods such as grid search and cross-validation,and finally the optimal prediction results of each model are output.Through comparative analysis,it is found that the errors of the LSTM model are the smallest and the prediction accuracy is higher;while the prediction accuracy of the Light GBM model is slightly lower than the LSTM model,but the training speed is fastest and the prediction effect is also better.In the end,in view of the shortcomings of the single model in the prediction process,the paper established passenger flow prediction models based on linear weighted fusion and Stacking fusion,respectively,based on the XGBoost,LightGBM and LSTM models,and compare the prediction result of the fusion model with the single model.The results show thatthe average absolute percentage errors of linear weighted fusion and stacking fusion models are 12.92% and 10.43%,respectively.Compared with the single model,the accuracy has been significantly improved,while also verifying the effectiveness and superiority of the fusion model.