Week 03: Trip Generation (Fundamentals)

CEE 6505: Transportation Planning

Department of Civil and Environmental Engineering
Islamic University of Technology (IUT), Bangladesh
Content
• Fundamentals of Trip Generation
• Modeling Methods
– Growth Factor
– Cross-Classification
– Regression
• Stability and Updating of Trip
Generation Models
Transportation Forecasting
• Transportation Forecasting is used to estimate the
number of travelers or vehicles that will use a given
transportation facility in the future
• The conventional approach to transportation
forecasting is based on what is commonly know as the
‘four step model’

3
Transportation Forecasting – 4 Step Model

• Trip Generation
– Estimates the number of trips from given origins and destinations
• Trip Distribution
– Determines the destination for each trip from a given origin
• Mode Choice
– Determines the mode choice for each trip
• Route Assignment
– Determines the specific route for each trip

4
Fundamentals of Trip Generation
• Trip Generation model is used to estimate the number of
person-trips that will begin or end in a given traffic analysis zone

The unit of analysis for traffic generation is the TAZ 5

Fundamentals of Trip Generation
• How many trips per zone?
• How many trips per family?
• Surveys to establish
– Number of trips as a function of number of people per
household, number of cars, type of dwelling, residential area…
– Distribution among trip purposes
– Distribution between motorized and non-motorized
– Distribution between chained and unchained trips
– Number of captive riders, e.g., f(number of people per
household vs. number of automobiles in the household)
6
Fundamentals of Trip Generation
• Trip or Journey
– A one-way movement from a point of origin to a point of destination
– Classically, only vehicular trips were of interest. However, now non-
motorized trips are also being considered, e.g., walking trips more
than 300 m or 3 blocks
– Infants under 5 can often be ignored

7
Fundamentals of Trip Generation
• Home-based (HB) Trip
– the home of the trip maker is either the origin or the destination of
the journey
– For visitors, hotel can be considered as temporary home

• Non-home-based (NHB) Trip

– neither end of the trip is the home of the traveler

8
Fundamentals of Trip Generation
• Trip Production
– the home end of an HB trip or as the origin of an NHB trip
• Trip Attraction
– the non-home end of an HB trip or the destination of an NHB trip

9
Fundamentals of Trip Generation
• Home Zone – the zone in which the trip maker lives
• Origin Zone – the zone where the trip starts
• Destination Zone – the zone where the trip ends
• Production Zone
– home zone for home-based trips
– origin zone for non-home-based trips
• Attraction Zone
– non-home-zone for home-based trips
– destination zone for non-home-based trips 10
 Fundamentals of Trip Generation
• Characterization of Journeys
– By Purpose
• a better understanding of travel and trip generation models can be obtained
if journeys by different purposes are identified and modelled separately

• Categories of HB trips:
– travel to work;
– travel to school or college (education trips);
Mandatory
– shopping trips;
– social and recreational journeys;
– escort trips (to accompany or collect somebody else); Optional
– other journeys (hospital, bureaucratic, personal business, etc.)

• NHB trips are sometimes separated into ‘on business’ and ‘other’ but are
often kept as a single category because they only amount to 15–20% of all
total travel 11
Fundamentals of Trip Generation
• Characterization of Journeys (contd…)
– By Time of Day
• Peak (AM or PM) and off-peak
• proportion of journeys by different purposes usually varies greatly with time of day
• Gets more complicated when tours rather than trips are of interest

– By Person Type
• individual travel behavior is heavily dependent on socioeconomic attributes
• Categories: income level, car ownership, household size and structure
• Should not have too many strata for each category.
12
Fundamentals of Trip Generation
• Factors Affecting Trip Generation
– Personal Trip Productions
• income;
• car ownership; Most common for household
• family size; trip generation studies
• household structure;

• value of land; Zonal studies

• residential density;

• accessibility. offers a way to make trip

generation elastic (responsive) to
changes in the transport system
13
Fundamentals of Trip Generation
• Factors Affecting Trip Generation (contd…)
– Personal Trip Attractions
• roofed space available for industrial, commercial and other services
• zonal employment
• Accessibility measures (very new)
– Freight Trip Productions and Attractions
• At most 20% in general even for industrial cities
• Contribution is significant in terms of congestion
• Important variables include
– number of employees
– number of sales
– roofed area of firm
– total area of firm
• No study included type of firm as variable (as per textbook!)

14
Production-Attraction vs. Origin-Destination
A worker leaves Zone 1 in the morning to
go to work in Zone 8

This results in 2 trip ends:

• One Origin for Zone 1
• One Destination for Zone 8

When that same worker leaves Zone 8 in

the evening to go to home to Zone 1

This results in another 2 trip ends:

Total Number of Trip Ends
• One Destination for Zone 1
• One Origin for Zone 8
Zone 1: 2 Trip Ends (1 O, 1 D)

Zone 8: 2 Trip Ends (1 O, 1 D)

15
Production-Attraction vs. Origin-Destination
A worker leaves Zone 1 in the morning to
go to work in Zone 8

This results in 2 trip ends:

• One Production for Zone 1
• One Attraction for Zone 8

When that same worker leaves Zone 8 in

the evening to go to home to Zone 1
Total Number of Trip Ends
This results in another 2 trip ends:
Zone 1: 2 Trip Ends (2 Productions) • One Production for Zone 1
• One Attraction for Zone 8
Zone 8: 2 Trip Ends (2 Attractions)
16
Production-Attraction vs. Origin-Destination
• Based on the convention of trip generation models
– Origins and Destinations are defined in terms of the direction of the trip
– Productions and Attractions are defined by the land use
• Residential Land use PRODUCES trip ends
• Non-residential land use ATTRACTS trip ends
• This is a useful distinction because of how trip generation models are
typically developed

17
Production-Attraction vs. Origin-Destination
A traveler living in zone 101 makes the following three trips (trip
numbers are in bold)

18
Production-Attraction vs. Origin-Destination

Trips 1,2, and 3 are made by a traveler living in zone 101

Trips 4 and 5 are made by a traveler living in zone 150
Trips 6, 7, 8, and 9 are made by a traveler living in zone 140
19
Trips 10, 11, and 12 are made by a second traveler living in zone 140
Production-Attraction vs. Origin-Destination

The objective of trip-generation step

is to come up with these numbers

20
 Production-Attraction vs. Origin-Destination
For example, Trip Ends for Zone 1 would be reported as Residential
1. 1000 Production Trip Ends
1
2. 500 Attraction Trip Ends

Non-residential

Trip generation models typically model separately, i) residential trip production, ii) non-
residential trip attractions

This approach works for home based trips (HB). But falls apart when we start to consider
non-home based trips (NHB). Special techniques are developed to deal with the
relatively small number of NHB that occurs.
21
Modeling Trip Generation

The trip generation model typically can take the form of

No. of trips = f(pop, income, auto ownership rates)
The model is developed and calibrated using BASE year data 22
Fundamentals of Trip Generation
• Trip Generation models are often developed from travel surveys. These
surveys are used to determine the trip making pattern for a sampling of
house holds in the area. This trip making pattern is then related to land
use (nominally) and socioeconomic factors that are considered to affect
travel patterns
• Often separate predictions are mode for different type of trips since travel
behavior depends on trip purpose
• In other words different models must be developed for each trip type
23
Modeling Trip Generation
• Modeling Trip Productions
Thp= Number of trips made by household h for purpose p
=fp[characteristics of household h, land-use, & transportation system]
and,
Pi= Number of trips produced in zone i =
Here, p= home-based work, home-based other, non-home-based, etc.
NOTE: The above equation will directly give the number of trips produced in zone
i only if there are no non-home-based trips, else additional processing is required
(will discuss this later on) 24
Modeling Trip Generation
• Modeling Trip Attractions
Aip= Number of trips of purpose p attracted by zone i
=fp[land use of the zone i]
= (for example)
= 2.0 X (Retail employment in zone i) + 1.7 X (Service
employment in zone i)
Ai= Number of trips attracted in zone i =
25
Modeling Trip Generation – Growth Factor
• Future trips can be calculated as:

• Where Ti = future trips in Zone i; ti = current trips in Zone i, and Fi =

growth factor which can typically be calculated as:

• Where P = population, I = income, C = car ownership; d = design

year and c = current year
• Normally used to predict future external trips only as they are few
in number and very hard to calculate rationally 26
Modeling Trip Generation – Growth Factor
Example 4.1 Consider a zone with 250 households with car and 250
households without car. Assuming we know the average trip generation rates
of each group:

car-owning households produce: 6.0 trips/day

non-car-owning households produce: 2.5 trips/day

Let us also assume that in the future all households will have a car; therefore,
assuming that income and population remain constant (a safe hypothesis in
the absence of other information).

Estimate the number of trips in future.

27
Modeling Trip Generation – Cross-Classification
The Classical Model
• Cross-classification is a technique for trip generation that uses categories or cells in
which the changes in one dependent variable (trips) can be measured when the
changes in other independent variables (lane use, socioeconomic status) are
analyzed
• Assumptions: trip generation rates are relatively stable over time for certain
household stratifications
• Limitations:
– heavily depends on census data and hard to forecast the number of households in each
stratum in the future
– Large amount of data are needed as the rates are found empirically.
• Got introduced in late 1960s in the UK and became preferred over regression
models very soon. 28
Modeling Trip Generation – Cross-Classification
The Classical Model (Household Based Approach)
• Average number of trips with purpose p can be calculated as:

• Where
– t = avg. number of trips
– p = purpose
– h = household type
– T = total number of trips
– H = total number of households 29
Modeling Trip Generation – Cross-Classification
The Classical Model (Household Based Approach)
• trip productions with purpose p by person type q in zone i, Oiqp

• Where
– i = zone number
– a = number of households

30
Modeling Trip Generation – Cross-Classification
The Classical Model (Household Based Approach)
• Predicting future number of households
– If probability distribution for income (I), car ownership (C) and household
structure (S) for a specific type of household is

– Then, the future number of households of this category is

– Where, Hi = total number of households in the zone i 31

Modeling Trip Generation – Cross-Classification
Total Home-Based-Non-Work Trip Rates
Persons per Household
Type of Area Vehicles per HH 1 2,3 4 5+
High Density 0 0.6 2.1 4.6 7.0
1 1.5 3.0 5.5 7.9
2+ 1.8 3.4 5.9 8.3
Low Density 0 1.0 2.5 5.0 7.4
1 1.9 3.5 6.0 8.4
2+ 2.3 3.9 6.4 9.0 32
Modeling Trip Generation – Cross-Classification
Number of Households in Target Year
Persons per Household
Type of Area Vehicles per HH 1 2,3 4 5+
High Density 0 100 200 100 100
1 200 300 200 100
2+ 100 200 100 200
Low Density 0 50 100 100 100
1 100 200 100 100
2+ 100 100 100 10 33
 Modeling Trip Generation – Cross-Classification
Number of Households in Target Year
Persons per Household
Type of Area Vehicles per HH 1 2,3 4 5+
High Density 0 60 420 460 700
1 300 900 1100 790
2+ 180 680 590 1660
Low Density 0 50 250 500 740
1 190 700 600 840
2+ 230 390 640 90
Number of Trips = trip rate*no. of HH = 0.6 * 100 = 60 Total trip = 13,080 34
Modeling Trip Generation – Cross-Classification

35
 Modeling Trip Generation – Cross-Classification
Example: Below is the summarized result of a survey.
Also, in another area of similar
characteristics, following information was
found from the census.

Calculate the total number of trips generated from this area.

36
Modeling Trip Generation – Cross-Classification
Solution:

37
Modeling Trip Generation – Trip Rate Analysis
• Trip-Rate Analysis
– Trip rate is estimated on characteristics of the trip generators with in
the zone. Production rates are determined using the characteristics of
the residential land uses and attraction rates using the characteristics
of the nonresidential land uses
Example
The characteristics of the trip generator is given in 1000 SQ. FT.
And the trip generation rate for each generator is given as TRIPS PER 1000 SQ. FT.
For example
Residential: Total 1000 Sq. Ft. = 2744 1000 sq. ft., Trip Gen. Rate = 2.4 trips/1000 sq.ft
TOTAL NO. of TRIP from residential land use = 2744*2.4 = 6586 Trips

This method of trip generation is often used to do site impact studies 38

Modeling Trip Generation – Cross-Classification
The Classical Model (Person Category Approach)
• Advantages
– A person-level trip generation model is compatible with other components of the
classical transport demand modelling system, which is based on trip makers rather than
on households
– In its cross-classification scheme it uses all important variables and yields a manageable
number of classes allowing class representation to be forecast more easily
– Requires lower sample size
– Demographic changes can be more easily accounted for
– Person categories are easier to forecast than household categories
• Disadvantages
– difficulty of introducing household interaction effects and household money costs and
money budgets into a person-based model
39
Modeling Trip Generation – Cross-Classification
The Classical Model (Person Category Approach)
• Formation

40
Modeling Trip Generation – Regression Analysis
The Basics
• Regression is a mathematically based procedure
• In trip generation step, no. of trip ends are treated as dependent variables,
which is a function of one or more independent variables
• Pi = f(x1, x2, x3, …, xn)
• Aj = f(k1, k2, k3, …, kn)
• Where
– Pi = trip productions from zone I
– Aj = trip attractions from zone j
– xi = predictive factors for productions (independent variables)
– ki = predictive factors for attractions (independent variables) 41
Modeling Trip Generation – Regression Analysis
The Basics
• Example: Pi = a + bx1 + cx2 + dx3
• Where
– Pi = trip productions from zone I
– x1 = number of car ownership
– x2 = family income
– x3 = no. of population
– a, b, c, d – model parameters to be calculated by regression analysis
• Model parameters and variables vary from one study area to another and
are established by using base-year information.
• Once the equations are calibrated, they are used to estimate future travel
42
for a target year
Modeling Trip Generation – Regression Analysis
Forms of Regression Models
• Simple linear (1 dependent & 1 independent variables)
– Pi = a + bx1
• Multiple linear (1 dependent & more than one independent variables)
– Pi = a + bx1 + cx2 + dx3
• Non-linear Regression
– Y = eax+b
– Y = ax2 + bx + c
• Zero-Intercept Cases
– Pi = ax1
– Pi = bx1 + cx2 + dx3 43
Modeling Trip Generation – Regression Analysis
Linear Regression
• Four guidelines are helpful in deciding which independent variables to
include in a linear regression model. The selected explanatory variables:
– Must be linearly related to the dependent variable
– Must be highly correlated with the dependent variable
– Must not be highly correlated between themselves
– Must lend themselves to relatively easy projection
• yi = a +bxi
• Calculate a and b as:

44
Modeling Trip Generation – Regression Analysis
Linear Regression (Correlation)
• The goodness of fit a regression line increases with the proportion of the
total variation that is explained by the line. The coefficient of
determination, R2 quantifies this fact
• It ranges from zero when none of the total variation is explained by the
regression line to unify when all the variation is explained by the line
• It is denoted as a squared quantity to capture the fact that it is always
nonnegative
• The square root of the coefficient of determination is called the coefficient
of correlation

45
Modeling Trip Generation – Regression Analysis
Linear Regression (Correlation)
• Its value can range from -1 to +1

• In the case of linear regression, the sign of r is the same as the sign of the
slope b of the regression line

• Figure on the next slide illustrates that r is near +1, there exists a high
positive correlation

• If it is near -1, there exists a high negative correlation

• And if it is around zero, there exists no correlation between X and Y. 46

Modeling Trip Generation – Regression Analysis
Linear Regression (Correlation)

47
Modeling Trip Generation – Regression Analysis
Example: Let the trip rate of a zone is explained by the household size done
from the field survey. It was found that the household size are 1, 2, 3 and 4.
The trip rates of the corresponding household is as shown in the table below.
Fit a linear equation relating trip rate and household size.

Answer:

48
Modeling Trip Generation – Regression Analysis
Zonal-based Multiple Regression
• finds a linear relationship between the number of trips
produced or attracted by zone and average socioeconomic
characteristics of the households in each zone
• Important considerations:
– Explaining variations
• Zonal models can only explain the variation in trip making behavior between zones
• Need inter-zonal variations adequately reflect the real reasons behind trip
variability
• Zones need to have homogeneous socio-economic composition
• Should represent a range of conditions as wide as possible
• Major concern: main variations in person trip data occur at the intra-zonal level 49
Modeling Trip Generation – Regression Analysis
Zonal-based Multiple Regression
• Important considerations: (contd…)
– Role of intercept
• Large intercept values are not good (in comparison to the product of the average
value of any variable and its coefficient)
• If the intercept is not significantly different from zero then re-estimat forcing it to
go through origin.
– Null zones
• It is possible that certain zones do not offer information about certain dependent
variables (e.g. there can be no HB trips generated in non-residential zones)
• Null zones must be excluded from analysis
50
Modeling Trip Generation – Regression Analysis
Zonal-based Multiple Regression
• Important considerations: (contd…)
– Zonal totals versus zonal means
• Zonal totals – trips per zone and cars per zone
• Zonal means - trips per household per zone and cars per household per zone
• Results are very similar but do not mix these methods in one model.
• inter-zonal variability diminishes with zone size

51
Modeling Trip Generation – Regression Analysis
Zonal-based Multiple Regression
• Important considerations: (contd…)
– Zonal totals versus zonal means
• Zonal totals

• Zonal means

• Where,

• And, Hi = number of households in zone i

52
Modeling Trip Generation – Regression Analysis
Household-based Regression
• In a household-based application each home is taken as an
input data vector in order to bring into the model all the range
of observed variability about the characteristics of the
household and its travel behavior
• Addresses these drawbacks of zonal based regression
– Smaller zone size can reduce intra-zonal variations but increased
number of zones will:
• Increase project expense (data collection, calibration and operation)
• Increase sampling errors (assumed non-existent by the multiple linear
regression model) 53
Modeling Trip Generation – Regression Analysis
Household-based Regression
• Example 4.3 Consider the variables trips per household (Y), number of workers (X1) and
number of cars (X2). Table below presents the results of successive steps of a stepwise model
estimation; the last row also shows (in parenthesis) values for the t-ratio (equation 4.9).
Assuming large sample size, the appropriate number of degrees of freedom (n − 2) is also a
large number so the t-values may be compared with the critical value 1.645 for a 95%
significance level on a one-tailed test (we know the null hypothesis is unilateral in this case as
Y should increase with both X1 and X2). Comment on the models.

54
Modeling Trip Generation – Regression Analysis
Household-based Regression
• Example 4.3 (contd…) The results of estimated and actual trips per household
(observed/estimated)

Stepwise regression method is not encouraged. Try all at a time approach 55

Modeling Trip Generation – Regression Analysis
Household-based Regression
The problem of non-linearity
• apparently linear relations may turn out to
be non-linear when the presence of other
variables is allowed for in the model

56
Modeling Trip Generation – Regression Analysis
The problem of non-linearity
• Two solutions available:
– Transform the variables in order to linearize their effect (e.g. take
logarithms, raise to a power)
– Use dummy variables
• the independent variable under consideration is divided into several
discrete intervals and each of them is treated separately in the model
• Linearity assumption is not necessary
• (n − 1) dummy variables are needed to represent n intervals
• May make forecasting about some variables easier

57
Modeling Trip Generation – Regression Analysis
The problem of non-linearity
Example Consider the model of Example 4.3 and assume that variable X2 is replaced
by the following dummies:
• Z1, which takes the value 1 for households with one car and 0 in other cases;
• Z2, which takes the value 1 for households with two or more cars and 0 in other
cases.
It is easy to see that non-car-owning households correspond to the case where both
Z1 and Z2 are 0.
The model of the third step in Slide 54 would now be:

58
Graphically test the linearity of the relationships.
Modeling Trip Generation – Regression Analysis
The problem of non-linearity
Example (contd…)

• would it not be preferable to

estimate separate regressions for
the data on each group, as in that
case we would not require each
line to have the same slope (i.e.
the coefficient of X1)?

59
Modeling Trip Generation – Regression Analysis
Obtaining Zone Totals
• Only required for household based models as zonal based models are as
they are estimated specifically at this level.
• Easy for linear models but complex for non-linear models.
• For the 3rd equation in Slide 54

• Where,
– Ti = total number of HB trips in zone i,
– Hi = total number of households in it, and,
– Xji = average value of variable Xj for the zone 60
Modeling Trip Generation – Regression Analysis
Obtaining Zone Totals
• For the equation in Example 4.4

• Where,
– Hji = number of households of class j in zone I
• For the model on previous slide, it would be necessary to estimate
the average number of workers per household (X1) for each car-
ownership group in each zone, and this may be complicated for
long-term forecasts 61
Modeling Trip Generation – Regression Analysis
Matching Generations and Attractions
• Fact: Regression does not guarantee that:
• Problem: this equation required by the trip distribution. (It is impossible to
have a trip distribution matrix where the total of trips obtained by sum of
all rows is different to that obtained by sum of all columns.)
• Solution:
– normally the trip generation models are far ‘better’ (in every sense of the word) than
their trip attraction counterparts
– The trip attraction models are at best estimated using zonal data. Hence, the total
number of trips arising from summing all origins Oi is the correct figure for T. So, all
destinations Dj are multiplied by a factor f given by:

62
Modeling Trip Generation – Regression Analysis
Stability and Updating of Trip Generation Parameters
• Temporal Stability
– remain constant over time (difficult to examine: panel data analysis)
• Geographic Stability (transferability)
– Suggest the existence of certain repeatable regularities in travel
behaviors which can be picked up and reflected by the model.
– Indicate a higher probability that temporal stability also exists.
– Allow to reduce substantially the need for costly full-scale
transportation surveys.

63
Modeling Trip Generation – Regression Analysis
Stability and Updating of Trip Generation Parameters
• Bayesian Updating of Trip Generation Parameters
– Formation

• Updated trip generation parameters

64
Modeling Trip Generation – Regression Analysis
Stability and Updating of Trip Generation Parameters
• Example 4.6 The mean trip rate, its variance and the number of observations for two
household categories, obtained in a study undertaken 10 years ago are shown below:

• New data:

65
 Thank You!

• Contact me
– Email: moinul@iut-Dhaka.edu