The Prediction Of The Dwell Time For Railway Freight Vehicles At The Start Station Based On Combining Model

Railway is a very important part of China's transportation system.With the rapid development of science and technology in recent years,many new modes of transportation have emerged in transportation market.And the observability of the trip is deeply loved by people.In the huge transportation market,railway freight has the characteristics of rich transportation and can carry large quantities of goods,but its punctuality is difficult to guarantee,which makes its competitiveness in the market weaker and also wastes railway resources Therefore,it becomes particularly important to study the deadline of railway freight vehicles,and the prediction of the dwell time of railway freight vehicles at the starty station is an important aspect of the study.First of all,through the research of the related content at home and abroad,find out the existing problems in the current research.Based on the existing problems,this paper studies the following four aspects:(1)the calculation of the historical dwell time of railway freight vehicles and data preprocessing: the railway freight information system has accumulated a lot of relevant data over the years,although the data types are rich and the number is huge,but It is necessary to select,clean and calculate the required data according to the research content to be applicable to the prediction of delivery time limit.Because the original data contains the message data of all stations in the whole railway network,it is necessary to first screen out the relevant message data of the originating station,and then eliminate the abnormal values to calculate the historical residence time through the message data.Then the influencing factors are analyzed and quantified.(2)Based on the BP neural network model,the freight vehicle dwell time at the start station is predicted: BP neural network is a commonly used artificial neural network.For the research content of this article,BP neural network can extract the non-linear part of the original data very well.You can also circumvent this drawback to a certain extent.By setting an appropriate learning rate,you can adjust the learning to an appropriate level of precision to continuously build the model more accurately.The experimental results show that the model has a good prediction effect.(3)based on the optimized SVR-GWO algorithm,the freight vehicle dwell time at the start station is predicted: The research of BP neural network is mainly aimed at the nonlinear part of the data.For the linear part,the SVR-GWO algorithm is used.The optimized SVR algorithm solves the shortcomings of traditional algorithms that are difficult to find the optimal parameters.The algorithm seeks the optimal penalty coefficient and other parameters in SVR,and then models it.Experiments show that the optimized SVR-GWO algorithm has better prediction effect on the residence time of freight vehicles at the departure station than the standard SVR algorithm.(4)based on the combining model algorithm,the freight vehicle dwell time at the start station is predicted: Based on the above research of BP neural network algorithm and SVR algorithm,each has its own characteristics.In order to give full play to the advantages of each algorithm,a combining model is proposed,which combines the nonlinear prediction of BP neural network with the linear prediction of SVR algorithm to form an optimal algorithm model,so as to more accurately predict the dwell time of railway freight vehicles at the start station.The experimental results show that the combining model composed of the weights determined by the reciprocal method of the error variance mean square has a better prediction effect.