Tickets Dynamic Allocation For Passenger Railway

With the gradual expansion of China's railway network scale and the increase of operation mileage,the capacity of passenger transport has been improved,and the phenomenon of insufficient passenger transportation in some areas has been alleviated.However,due to the uneven distribution of population and urban density,the travel demand of railway passengers in the road network are different,leading to the phenomenon of "empty seats on the train" and "no ticket" at stations.This thesis is from the perspective of the railway passenger marketing decision system.It analyzes the critical influence factors of the passenger tickets allocation from supply and demand,respectivily.A series of research about passenger ticket assignment was developed under condition of the railway network.The mainly contents include the following five items:(1)Using the system theory,the issue of tickets allocation is taken as the vital problem for realizing dynamic matching between the passenger railway supply and the passenger demanding,based on the railway passenger marketing system as the foundation.It presented two-phase ticket allocation method under the network environment.Moreover,it analyzed the complexity,time-varying and the supply and demand diversity of the ticket resources.(2)The retrospective data mining method was applied to analyze the importance features of train characteristic that affecting the train capacity utilization for any passenger line.Firstly,a seris of evaluation indexes are described for the train capacity utilization.Then,in terms of the train capacity utilization,the comparison and analysis are carried out for the two approaches: predictive optimization modeling and the retrospective data mining.Furthermore,taking Beijing Shanghai high-speed operation data as an example,PCA is used to analyze the train characteristic vectors with multi dimension which are the sample elements.As the mainly index for the train capacity utilization,the passenger load factor and other principal components are clustered.Finally,the effective features could be found by using the combined analysis method iteratively.(3)With respected to the analysis of the ticket purchasing behavior characteristics,a double-class clustering model for passenger classification was constructed based on passenger travel intensity,and the resolve algorithm is given.Firstly,the two commonly used methods of passenger selection are summarized,which are respectively supervised and unsupervised analysis methods.Secondely,based on the real data from Guiyang Guangzhou high-speed railway,combined with the ticket purchasing trend,we analyzed the critical effective factors of ticket purchasing behavior from three aspects: passengers travel factors,ticket purchasing factors and individual factors.At last,the fuzzy C-means double-class clustering model was constructed based on the passenger travel intensity to achieve the different passenger classification considering the ticket purchasing behavior.(4)According to the different travel schemes under the network condition,a transfer service network is constructed to be applied to build the passenger flow assignment model for trains considering stochastic user equilibrium.First of all,the passenger travel mode is analyzed through the railway passenger flow structure.And based on the network structure of train service,a transfer node is represented by multiple nodes to help to form the transfer service network.Then,according to the seasonal trend rule of passenger trip distribution,time series analysis method is used to partition passenger flow mode by different seasons,and get prediction model with ARMA to modal identification.Moreover,under the condition of stochastic user equilibrium,a passenger flow assignment model is constructed based on the prediction model and the transfer service network.Thus,the effective passenger flow is allocated to the trains in the transfer service network.And the gradient projection algorithm is applied to solve the problem.Finally,the model is verified by an example.(5)With the passenger train allocation results,we formulate the sectional average ticket intensity function based on support vector machine and put forward the dynamic ticket allocation model based on semi-Markov decision process combining the sectional average purchasing intensity function.Firstly,the prediction function of sectional average ticket intensity is constructed by using the nonlinear support vector machine and the time variable which is pre-sale date.Then,with the sectional average ticket intensity function,the semi Markov decision process is applied to describe the daily ticket pre-sale process.In the last,the dynamic ticket allocation model is constructed for maximizing the expectation revenue.And the validity of the model is verified by comparative analysis with the traditional ticket organization.