Department of Transportation and

Geotechnical Engineering, CoET

TR 631: Highway Planning and Economics

Topic 2.3 – Other Methods

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Since individuals choose whether to make specific trips,
discrete choice models such as binary logit can be used
to predict trip productions.
 With binary logit, the probability that an individual will
choose to make one or more trips (as opposed to not
travelling) can be expressed as

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 From the estimated coefficients, you can see how the
explanatory variables will impact the probability with
which an individual will make a certain trip.

 In addition, you can aggregate the disaggregated

probabilities to obtain the proportion of the population
that will take this type of trip, and thus generate the
aggregate number of trips produced by a zone.

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Interpreting the results of a logit model

 A logit model predicting trip making probability yielded

the following results

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Interpreting the results of a logit model

 With a goodness-of-fit ('Adjusted Rho Squared') of 0.22

the variables Sex, Age, Married Female, Married Male,
Fem w/ Child < 6 are all dummy variables (i.e. equal to
1 or 0).

 Note that all of the coefficients are significant at a 95%

confidence level (t-statistic > 2).

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Interpreting the results of a logit model
 This logit model predicts the probability with which an
individual will make a work trip according to the
following equation:

 From the estimated coefficients, you can see how the

explanatory variables will impact the probability with
which an individual makes a trip to work.

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Interpreting the results of a logit model
 For example, the coefficient for education (0.211)
suggests that, all else being equal, people with more
education are more likely to make work trips than those
with less education.

 Note that none of the signs from the model seem

unreasonable.

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Interpreting the results of a logit model
 You can also use the equation above to calculate how a
change in an explanatory variable will impact a person’s
probability of making a work trip.

 For example, a person with a high school diploma (Educ

= 10) who has a 50% probability of making a work trip,
would, all else being equal, have a 70% probability of
making a work trip if he or she had a Bachelor’s Degree.

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Example
 The Dutch National Transport Model System is fully
based on disaggregated discrete choice models of
travel demand.

 The trip frequency submodel predicts the probability that

an individual with known characteristics will make no,
one or more round tours for specific trip purposes on an
average day.

 In later stages, tours are decomposed into trips.

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Example
 The unit of analysis is a person, member of a household
with known characteristics.

 This trip production model distinguishes 5 trip purposes,

further divided by person type.

 In total, eight different models are distinguished.

 The model has a two-stage structure:

o In step 1, the model predicts the probability of
making no versus one or more tours, thus a binary
choice model.
CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Example
o In step 2, another binary choice model predicts the
conditional probability of making one tour versus two
or more tours given the fact that tours are being
made.

o This process is repeated several times.

 Model 2 is applied consecutively in the cascade.

 This second model is identical in all these steps (but is

different from model 1).

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Example
• Structure of disaggregated trip choice models for trip production
in the Dutch National [Source: LMS, 199?]

Transport Model System

pp

CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Example
 It is a so-called stop/repeat-model.

 Both models are binary logit models with linear utility

functions including dummy variables describing
personal and household characteristics.

 In practical applications, instead of working with

probabilities, the expected number of tours is
calculated.

 This expected number of tours (ENT) per purpose may

be calculated using:
CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Example
 This expected number of tours (ENT) per purpose may
be calculated using:

where Pr(1+) is the probability outcome of model 1, and

Pr(R) is the probability of making additional trips which
is the outcome of model 2.


CoET
Department of Transportation and Geotechnical Engineering
Discrete Choice Methods
 Example
• Structure of disaggregated trip choice models for trip production
in the Dutch National [Source: LMS, 199?]

Transport Model System

pp

CoET
Department of Transportation and Geotechnical Engineering
Trip Generation with Count Data Models
 Poisson model

CoET
Department of Transportation and Geotechnical Engineering
Artificial Neural Networks (ANN)
 The classical transportation planning methods are
based on simple extrapolation of trends.

 Some mathematical models like linear regression

models have also been used by researchers for
estimating traffic generation for future period.

 However conventional models do not account for

nonlinearly in the model.
 ANN models account for nonlinear relationship
between independent variables and the dependent
variables and can be used to estimate traffic
production and attraction.
CoET
Department of Transportation and Geotechnical Engineering
A Fuzzy Expert System
 It can be used to estimate number of trips generated by
each traffic zone.

 The input of this system is inexact and inexplicit human

knowledge in form of facts and certain or uncertain
rules.

 In other words, having unclear information for future

values of independent variables, it is possible to gain an
insight to the dependent variable’s value for the horizon
year.

CoET
Department of Transportation and Geotechnical Engineering
A Fuzzy Expert System
 In this method the final output, i.e. the number of
generated trips for each traffic zone will be found using
a reasoning scheme.

 Therefore, the fuzzy expert system enables us to

provide the future traffic volumes, with a reasonable
precision and without the need of exactly predict the
future values of independent variables.

 A number of studies have devoted to traffic flow

modeling using this computational intelligence.

CoET
Department of Transportation and Geotechnical Engineering
Activity Based Modelling
• Activity-based modelling of travel demand belong to the
third generation of travel demand models, which received
extensive interest.

• It treats travel as being derived from the demand for

activity participation.

• Various methods used in activity-based travel analysis,

include discrete choice models, hazard duration models,
structural equation models, and computational process
models.

CoET
Department of Transportation and Geotechnical Engineering
Activity Based Modelling
• Discrete Choice Models
• Discrete choice models have been used to model
complex travel behavior by the activity-based approach
since 1980s, although it was originally developed and
applied in the context of trip based framework.

• Hazard Duration Models

• Hazard-based duration models have been applied to
travel demand modeling since 1980’s.
• The general idea of a hazard-based duration model was
that it tried to model the conditional probability of “failure”
at time t, given that failure had not occurred prior to this
time.
CoET
Department of Transportation and Geotechnical Engineering
Activity Based Modelling
• For example, one might try to model the probability that a
worker finds a job at time t, given that she/he is
unemployed up to this time.

• The most pertinent application of the hazard-based

duration in activity-based travel demand modeling is in
connection with modeling the activities and home-stay
duration.

• Another use of hazard-based duration models is

modeling the time until the next particular activity occurs.
• For example, the time between shopping activities could
be modeled with the appropriate data.
CoET
Department of Transportation and Geotechnical Engineering
Activity Based Modelling
• Structural Equation Models
• Structural equation models applied in travel demanded
modeling were first conducted by Golob (Golob and
Meurs 1987; Golob, 2003).

• Recently, more and more researchers started to apply

this methodology to include activity-based travel demand
modeling.
• For instance, Golob (2000) developed a trip generation
model to forecast three sets of endogenous variables: (1)
activity participation (2) travel time, and (3) trip generation
as a function of household characteristics and
accessibility indices.
CoET
Department of Transportation and Geotechnical Engineering
Activity Based Modelling
• Structural Equation Models
• The results showed that, the basic model, which had 10
endogenous time use and trip generation variables and
13 exogenous variables, fit well, and all postulated
relationships were upheld.

• Rule Based Simulation Models

• A number of rule based simulation models have been
developed since 1980s, CARLA , STARCHILD, SAMS
(developed by the late Kitamura in 1996).

CoET
Department of Transportation and Geotechnical Engineering
Activity Based Modelling
• Rule Based Simulation Models
• The AMOS model, which was at the heart of the SAMS
model, was an example of Rule Based Simulation Model
that had been applied to a real-world policy analysis
situation.

• The development and application of the AMOS model in

the Washington, DC area was designed to demonstrate
how an activity-based travel demand model could be
used to forecast commuters’ short-term responses to the
type of TCM measures.
• Other representative simulation models include
SCHEDULER (Garling et al, 1994).
CoET
Department of Transportation and Geotechnical Engineering
Activity Based Modelling
• Treat travel as part of set of wider household activities

• These are categorised by “need” or “function”:

o subsistence (e.g., work)
o maintenance (e.g., shopping)
o discretionary (e.g., recreation)

• Finite number of hours in a day/week.

• Households prioritise time spent on each activity:
o trip chaining
o trip substitution
• Structural equation models.
CoET
Department of Transportation and Geotechnical Engineering