Freight Products Design And Service Network Optimization In Railway Transport:Theory And Methods

To serve shippers with all types of commodities, railway companies should design differentiated and targeted freight products, according to the diverse demand characteristics of different shippers and market segments, in terms of personalized and heterogeneous preference. Within this context, based on demand analysis, this thesis proposed the products design models and methods for railway shippers with heterogeneous preference, and designed the theoretical framework for demand-oriented products design; then, based on the differentiated value of service held by various railway shippers, this thesis designed a railway service network optimization model, to solve the service plans design problem with multiple and diverse transport demands. In general, this thesis aims at the theory and methods related products design and service network optimization problem in railway transport within tactical level, which will update the service design strategy, and increase the competitiveness of railway freight services in freight markets.The outline of this thesis includes:Firstly, railway freight product is defined as displacement service from origin to destination and a bundle of service attributes. According to the demand characteristics in railway freight market, this thesis described the railway shippers' choice behavior pattern affected by both observable and unobservable variables, then introduced five different decision rules applied by shippers, including dominance, satisfaction, lexicographic rules, utility and regret. Besides, an introduction to the value of service and market segmentation strategies was provided.Secondly, transport price, time, frequency, reliability and safety are taken as service attributes of railway freight transport, to design the service profiles and choice sets with Choice based Conjoint (CBC) analysis, orthogonal design and shifting method. Three rules are applied to examine shippers' preference based on Multinominal Logit (MNL) model and Mixed Logit(ML) model, including Random Utility Maximization (RUM), Random Regret Minimization(RRM), and Hybrid RUM-RRM. Shippers values of service are described as value of time(VOT), value of reliability (VOR), value of frequency (VOF), value of safety (VOS). All coefficients are estimated based on the Maximization Likelihood Extimation (MLE) technique.Thirdly, two market segmentation methods are presented, i.e., exogenous market segmentation method with K-means cluster algorithm, and endogenous market segmentation method with Latent Class (LC) model and Latent Class Mixed Logit (LCML) model. The latter method is capable of measure shippers' preference heterogeneity discretely distributed across distinct classes, and continuously distributed within classes. Bayesian Information Criterion is used to detennine the best size of market segments. In general, three market segments, including price sensitive segment, service sensitive segment, and balanced price-service sensitive segment are elaborated in terms of products design strategies. Freight product with floating prices and basic service should aim at shippers within price sensitive segment,product with differentiated price and good service should serve shippers within service sensitive segment, and products with consistent price and service quality should be offered to shippers within balanced price-service sensitive segment. Besides, overall differentiated marketing and partially centralized marketing strategies are provided.Fourthly, the tradeoff and relationship between transport price and service are discussed with a theoretical model, to study the impact of value of service on the efficiency of railway service network optimization. Both value of service and demand volume should be taken as priority criteria when service links capacity is limited. Besides, value of service is an important cost component for the generalized cost of service network. In detail, a interger model was designed with the minimization of generalized cost including operation cost, transport cost,transshipment cost and service cost, subject to the link capacity constraint and transshipment constraint. Service plans are designed with K shortest path algorithm. Improvement plans are discussed for Simulated Annealing algorithm, and pseudo-code is presented for the improved Simulated Annealing algorithm. The whole steps to solve the interger model was provided.Finally, shippers' preference heterogeneity and value of service are estimated with RUM model, RRM model, and Hybrid RUM-RRM model, respectively, based on a case study in southwest of China. Coefficients for service attributes parameters are compared and discussed,to elicit how shippers will react to different services with various service levels, and how shippers will make tradeoff between price and service quality. Further study of market segmentation by Latent Class model shows that Hybrid RUM-RRM model preforms better in terms of model fit than RUM model, RRM model. The best size of market segments is defined by the value of BIC. Then,a series of services with differentiated service attributes and service levels are provided to shippers with multiple value of service and heterogeneous preferences,to find the best service selection, transshipment nodes, and service frequency, within a local railway network with 14 nodes and 20 links.Results of the service network optimization imply that single and less targeted products cannot meet the requirement of shippers with diverse and heterogeneous demand preferences.To optimize the generalized cost of railway service network, railway companies should design differentiated and targeted freight products, to offer freight services with different service price and service level. Results also reveal that the improved Simulated Annealing algorithm preforms better than traditional basic SA. Policy advices and suggestions are discussed for the product design in railway freight transport.