1、谭志加 Office #616 ,华中科技大学管理学院,交通组织运输学,The 4-Step Model,Transportation Planning,Modal choice,train,bus,auto,Transportation Planning,Transportation Planning,Modal Choice,Overview,MethodologiesDiscrete choice analysisLogit models,Transportation Planning,Mode usage (modal choice, or modal split) analysis
2、is to estimate the proportion of trip-makers traveling between each pair of origin-destination zones who are likely to use each mode (public transit or auto). Understand the factors that influence individual choices of travel modes Evaluate the impacts of alternative transport policies on modal spli
3、t How to achieve a balanced use of different travel modes through efficient traffic control and management (e.g. road pricing),Objectives of Modal Choice Analysis,Transportation Planning,Factors Influencing the Choice of Mode,Characteristics of the Trip-maker Car availability and/or ownership Posses
4、sion of a driving license Household structure Income Residential density,Characteristics of the Journey (Trip) The trip purpose Time of Day when the journey is undertaken,Transportation Planning,Factors Influencing the Choice of Mode,Characteristics of the Transport Facility Quantitative factors suc
5、h as: Relative travel time Relative monetary costs Availability and cost of parking Qualitative factors (less easy to measure) such as: Comfort and convenience Reliability and regularity,Transportation Planning,Basic Consideration of Modal Choice,The factors influencing the choice of modes are used
6、as independent variables to be included in mathematical models of modal choice. The dependent variable being the market share or the percent of travelers that are expected to use each of the available modes. The planner looks at how these characteristics interact to jointly affect the trip makers ch
7、oice of mode. When the relationships have been discovered, the planner can predict how the population of the future will choose from among the modes that will be available.,Transportation Planning,Utility and Disutility Functions,Modal choice has a relationship with the utility.The utility (or disut
8、ility) function is typically expressed as the linear weighted sum of the independent variables or their transformation:,V is the utility derived from a choice defined by the magnitudes of the attributes X that are present in that choice and that are weighted by the model parameters a, parameters can
9、 be estimated by linear regression, based on actual choice results.,Transportation Planning,Example of Disutility Functions,A linear utility function:,WhereX1 = in-vehicle travel timeX2 = monetary costX3 = cost/incomeX4 = reliability,Continuous vs. Discrete Goods,Bus,Auto,Indifferent curve,Budget cu
10、rve,Transportation Planning,Discrete Choice Analysis,Time cost,Monetary cost,Continuous vs. Discrete Goods,Transportation Planning,Discrete Choice Analysis,快递公司,Discrete Choice Framework,Decision-Maker-Individual (person/private/public sector)-Socio-economic characteristics (income, objectives) Alte
11、rnatives-decision-maker n selects one facing Jn alternatives set Cn=1,2,Jn with Jn alternatives Attributes of alternatives-Travel time, cost, reliability, safety. Decision rules-Dominance, satisfaction, utility etc.,Transportation Planning,Discrete Choice Analysis,A Simple Example,tastes,Transportat
12、ion Planning,Discrete Choice Analysis,If U(train)U(bus)- train Probability(train)=1 If U(train) bus Probability(train)=0,Transportation Planning,Discrete Choice Analysis,A Simple Example,Imperfect information,Transportation Planning,Discrete Choice Analysis,Random Utility Model,Random Utility Model,
13、Decision rule: Utility maximization-Decision maker n selects the alternative i with the highest utility Ui among J alternatives in the choice set C,Systematic utility: function of the observed variables,Random utility,Choice probability,Transportation Planning,Discrete Choice Analysis,Binary choice,
14、Transportation Planning,Discrete Choice Analysis,Random Utility Model,Binary Logit (Logistic Probability Unit),Transportation Planning,The Logit Model,Transportation Planning,The Logit Model,S-shaped logit curve used to fit the model data in the case of two modes. A share model to divide the persons
15、 between the various modes according to each modes relative desirability for any given trip. Modes are relatively more desirable if they are faster, cheaper, or have other more favorable features than competitive modes. The better a mode is, the more utility it has for the potential traveler.,Transp
16、ortation Planning,The Logit Model,Different in times, cost, etc.,Probability of using mode i, Pi, is given by,Sum corresponding to all competing modes,Utility -measure the degree of satisfaction of the modeDisutility -measure the generalized cost (impedance),if given disutility functions,Simplest ca
17、se,bus,auto,Binary Choice (Bus vs. Auto),Transportation Planning,The Logit Model,Generalized travel cost,I do not care the cost. For any travel cost, I choose one in probability .,I do care the cost very mush. I choose the one with lower cost in probability 1.,more sensitive,Binary Choice (Bus vs. A
18、uto),Transportation Planning,The Logit Model,Bus share (%),Bus more preferable,Limiting cases,What happens as,What happens as,Transportation Planning,The Logit Model,Binary Choice (Bus vs. Auto),How about the general case?,(1) What happens as,(2) What happens as,Transportation Planning,The Logit Mod
19、el,Assume that there are two alternatives for Hankou and Wuchang, i.e. Car and Train. Based on survey and calibration, the utility function for the two alternatives are given by,1. Applying logit model to determine the number of passengers using Train QTrain and the car volume between Hankou and Wuc
20、hang QCar (assume car occupancy is 1, i.e. one passenger par car).,2. Determine the car volume suppose, due to traffic congestion, the travel time by car will depends on the car volume on the bridges, given by,3. Change the toll charge for car (Toll), so that the car volume is reduced to,Example,Tra
21、nsportation Planning,1. Applying logit model to determine the number of passengers using Train QTrain and the car volume between Hankou and Wuchang QCar (assume car occupancy is 1, i.e. one passenger par car).,Solution,Transportation Planning,2. Determine the car volume suppose, due to traffic conge
22、stion, the travel time by car will depends on the car volume on the bridges, given by,Solution,Transportation Planning,3. Change the toll charge for car (Toll), so that the car volume is reduced to,Solution,Transportation Planning,Transportation Planning,Independence of Irrelevant Alternatives Prope
23、rty (IIA): The IIA property holds that for a specific individual, the ratio of the choice probabilities of any two alternatives is entirely unaffected by the systematic utilities of any other alternatives in the choice set.,The Logit Model,不相关选择项的独立性,Red/Blue Bus Example,Suppose in a corridor, V(car
24、) = V(bus), so, P(car) = P(bus) =1/2 Now suppose that the bus company pants half buses into blue color (the remaining half buses are still in red color). We thus have RED and BLUE buses as two of the available alternatives. Applying the Logit modelP(car)=1/3; P(red bus)=1/3; P(blue bus)=1/3 SoP(bus)
25、:1/2 2/3 (Impossible in reality),after changing color,Observed result: P(car)=1/2; P(red bus)=1/4; P(blue bus)=1/4 some problems?,Transportation Planning,Nested Logit Model,Hierarchical Choice Structure,Overcome the IIA Problem of Multinomial Logit Alternatives are correlated (e.g., red bus and blue
26、 bus),Transportation Planning,Nested Logit Model,Arunotayanun, K., 2010. Taste heterogeneity and market segmentation in freight shippers mode choice behaviour. Transportation Research Part E (in press).,Level of service for freight modal choice: Monetary cost Time cost Reliability Flexibility Qualit
27、y Cargo type and value,Transportation Planning,The Application of Discrete Choice Analysis,Example Suppose the dis-utility of the customers demanding the package delivery service between a specific OD pair is captured using service fare and time costwhere parameter is called value-of-time and follow
28、s the following cumulative distribution,For a certain commodity (e.g., weight below 100g), the total demand is exogenously given, Q.,Transportation Planning,The Application of Logit Model,Depicting the suppliers in the market,Transportation Planning,快递公司,The Application of Logit Model,Discussions (1
29、) Which supplier will survive in this market? Why? (2) How does the suppliers act to attract more demand? (3) How to calculate the market share for a certain supplier? (4) How to capture individual customers perceived error?,Transportation Planning,The Application of Logit Model,Service fare,Service
30、 time,(1) Which supplier will survive in this market? Why?,Transportation Planning,The Application of Logit Model,(2) How does the suppliers act to attract more demand?,Service fare,Service time,Transportation Planning,The Application of Logit Model,Service fare,Service time,(3) How to calculate the
31、 market share for a certain supplier?,Transportation Planning,The Application of Logit Model,Service fare,Service time,(4) How to capture individual customers perceived error?,Mixed logit model,Transportation Planning,The Application of Logit Model,Piyush, T., Hidekazu, I., Masayuki,D., 2003, Shippe
32、rs port and carrier selection behaviour in China: a discrete choice analysis. Maritime Economics & Logistics, Vol. 5, pp: 23-39.,Transportation Planning,The Application of Logit Model,Transportation Planning,The Application of Logit Model,Problem,Methodology,Standard multinomial logit Model,Port cha
33、racteristics: ship calls, total TEU, numbers of berths, cranes, offered routes Shipping line characteristics: scale (TEU, fleet size) Shipper characteristics: distance, trade type,Transportation Planning,The Application of Logit Model,Findings/Contributions,Significance analysis: the significant var
34、iables of port are distance, number of berths, TEU. The significant variables of carriers is fleet size. Elasticity Estimation (to capture market shares):,Transportation Planning,The Application of Logit Model,Verma, R., Pullman, M., 1998. An analysis of the supplier selection process. OMEGA int. J.
35、 Mgmt. Sci, Vol. 26 (6), pp: 739-750. Verma, R., Kimes, S. and Dixon, M. “New Applications of Customer Choice Modeling in the Hospitality and Related Service Industries”, Cornell Hospitality Quarterly, forthcoming, 2010.,Transportation Planning,The Application of Logit Model,Hints: Making choice from discrete alternative set (vendor, supplier choice process; carrier choice; route assignment),
