Study On Optimization Of Railway Short-Term Passenger Flow Forecasting Method Based On Machine Learning

Passenger flow forecasting is one of the important work of the railway passenger transport department,especially in the background of the rapid growth of railway transportation capacity,the change of supply and demand contrast and the reform of the railway enterprise market.From the macro perspective,the forecast of medium and long term passenger flow has important reference value to the construction of railway infrastructure,the planning of road network and the design of railway passenger transport products.From the microcosmic perspective,the short-term passenger flow forecast has great significance to help the railway departments to optimize the ticket organization strategy,improve the utilization rate of railway passenger transport capacity,and improve the railway transportation enterprises.In this context,a large number of research and prediction methods have been carried out around railway passenger flow prediction.However,the different forecasting methods have each applicable scene,so no prediction method can be applied to all the prediction scenes.Also,the prediction methods often contain various parameters,and the different parameter values will have a greater impact on prediction results.Therefore,it is worthwhile to study how to select the best prediction methods and parameters in the prediction scenario.To this end,based on the data of railway ticket selling,this paper develops short term passenger flow forecasts with specific departure dates,train trips,OD.The prediction effects of the five models in different scenarios are compared,and the influencing factors are analyzed.What's more,by extracting the relationship between the feature information in the prediction scene and its prediction effect,this paper puts forward the idea of using machine learning to optimize the prediction model.A classification algorithm based on machine learning is constructed and implemented,which can help railway business departments choose relatively optimal forecasting methods under different forecasting scenarios,to improve the effect of prediction.