Research On The Application Of Hybrid Model In Railway Freight Demand Forecasting

By studying the development history of the transport industry in developed countries in Europe and America,compared with other traditional independent transport of railway and water,the containerized multimodal transport is the best mode of transport for economic development.Container multimodal transport saves resources and reduces costs,and the utilization rate of container is high,the main public iron water transport information exchange is smooth,convenient service.China's container transport development space is large,railway container transport with a large number of capacity channels and the completion of 18 container logistics centers and high-speed development.However,there is a lack of container multimodal transport information system,and the lack of information exchange between transport entities leads to a high air running rate.The problem in this paper is to forecast the container traffic demand of the national railway(a certain region),which provides decision support for the dynamic deployment development of container stations in the future.By studying the railway container traffic data,it is found that the railway container traffic data has obvious time series characteristics and is greatly influenced by the goods output in the area where the shipping station is located.This article through to railway container traffic volume time series ARIMA model,and factor of influence in the National Bureau of Statistics web site to collect relevant data,through calculating the related influencing factors and railway container of Spearman rank correlation coefficient,obtain high degree of railway container traffic related influence factors of training as a forecast of railway container DBN input data and then,and is obtained by the improved particle swarm algorithm Sigmoid-OAPSO predict DBN model structure of railway container,By railway container ARIMA model and DBN model of hybrid model,and get the corresponding proportion in the hybrid model coefficient,at this time the mixture model on the RMSE of the minimum training data set,then sub models ARIMA and DBN coefficient do fitting,and predict coefficient of 15 months in the future,and on the test set inspection prediction accuracy.The innovations and contributions of this thesis are as follows:(1)An improved particle swarm optimization algorithm(Sigmoid-OAPSO)is proposed.Sigmoid random variable is introduced instead of uniformly distributed random variable to avoid local minima.The new velocity and new position of particles are calculated and the improved particle swarm optimization algorithm is obtained by combining three strategies.(2)A hybrid method of ARIMA and DBN model algorithm is proposed.ARIMA and DBN are both sub-models of the final prediction model,and their proportion coefficient in the hybrid model is also characteristic of time series through analysis.Therefore,ARIMA is fitted and tested on the test set,and it is proved that the hybrid model proposed in this paper has high prediction accuracy.(3)Through the analysis of the demand for the development of railway container intelligence and information,the demand analysis of railway container transport demand prediction system will be obtained according to the demand,and the development and implementation of the systemIn this paper,a comparative experiment is conducted to prove the accuracy and efficiency of the hybrid model,and to effectively solve the problem of container scheduling ratio in the future,so as to provide help for the intelligent and informationbased development of railway container transport.