Information Technology and Management Science

doi: 10.2478/itms-2013-0014
___________________________________________________________________________________________________________ 2013 / 16

Regression Analysis for Transport Trip Generation

Evaluation
Nadezda Zenina1, Arkady Borisov2, 1-2 Riga Technical University

Abstract The paper focuses on transportation trip generation For each type of generated trips, there are various
models based on mixed-use and transport infrastructure near the calculation methods (Fig. 2), but, as a result, the number of
site. Transport trip generation models are considered with an
trip attraction should be the same as the number of trip
aim to improve the accuracy of transport generated trips.
Information systems are reviewed, and smart growth criteria production.
that could affect the accuracy of trip generation models are also
identified. Experimental results of transport generated trips Input data: Input data:
based on linear regression equations and smart growth tools demographic and employment, land
are demonstrated. socioeconomic data use

Keywords Information systems, linear regression equations,

transport trip generation models Cross-classification Linear regression
models models

I. INTRODUCTION Trip correction:

Trip generation model is the first step of the classic four- 1. Local conditions
2. Mixed use
step transport model [10] trip generation, distribution, mode 3. Transport infrastructure
choice and trip assignment. Trip generation models are
divided into three levels: strategic, tactical and operational
ones. Cross-classification models or category analysis [10] Trip production = Trip attraction
[11] are used for trip generation calculations at the strategic
and tactical levels. Linear regression models [4] and Rate
methods with linear regression equations or with average rates Generated transport trips
are used to calculate transport generated trips at tactical and
operational levels [5], [18], [19].
The aims of the paper are to review Rate methods for trip
Fig. 2. Calculation process of generated transport trips
generation models and to evaluate smart growth tools to
increase the accuracy of generated trip calculations with
Rate methods with linear regression equations (1) are the
mixed-land use and transport infrastructure availability (public
most frequently used methods to evaluate the number of
transport, pedestrians). All calculations are performed at the
generated trips based on historical data. Transport generated
operational level and for trip attraction models.
trips are expressed as the number of trips per unit X, where
X the factor that describes the activity of land use, for
II. TRANSPORTATION TRIP GENERATION METHODS
example, for retail the gross leasable area, for residential
Transport trip generation is the process of estimating daily buildings the number of apartments.
or hourly person trips for an average weekday generated by Linear regression equations evaluate the number of
households. Transport generated trips include two types of generated trips that attract research area (dependent variable)
trips: trip production and trip attraction (Fig. 1). from independent variables

Trip production = Trip attraction Y = a + b1 x 1 + b 2 x 2 + + b n x n , (1)

where Y the dependent variable (trips/household), x1, x2

Pi Aj ... independent variables (population, number of apartments,
gross leasable area), b1, b2..bn regression coefficients that
show to what extent Y changes, if n variable increases.
The current paper considers various information systems for
trip generation calculation based on regression equations
Fig. 1. Types of transport generated trips. Trip production models estimate the and/or average rates.
number of person trips generated by each household for each of the trip
purposes and trip attraction models estimate the number of person trips
attracted to each location

89
 Information Technology and Management Science

2013 / 16 ___________________________________________________________________________________________________________

A. TRICS Information System different demographic factors, living conditions,

TRICS Information system [15] evaluates the transport economic factors;
capacity of a site on a street network based on the site different driving behaviours on roads (travel speed,
functional data (retail, residential, office spaces), location (in interval between cars).
the city centre or outside the city), analysed time period (peak In this study, the automobilization level was used as an
hour, working day, weekend) and multimodal transport additional correction for local conditions.
accessibility to an adjacent site. The main disadvantage of
TRICS is the insufficient number of studies for specific land III. ITE TRIP GENERATION INFORMATION SYSTEM AND
use categories that leads to low accuracy of generated trip SMART GROWTH TOOLS
calculations. Smart growth principles are strategies of transit-oriented
and pedestrian-friendly mixed use lands for generated trips by
B. NZTPD and RTA Information Systems
reducing the number of cars in communities. ITE trip
NZTPD (New Zealand Trip Rate and Parking Database) generation manual shows average daily and peak hours rates
[16] and RTA (Roads and Traffic Authority of New South for transport generated trips (Fig. 3). All data are mostly
Wales) [17] are widely used to estimate site impacts on provided for isolated areas without public transport and
transport infrastructure in New Zealand and the surrounding pedestrian infrastructure. In mixed-use conditions, the
area. In [9] several types of land use retail, residential, following steps are necessary to carry out for generated trip
offices, medical buildings with TRICS information system calculations:
were compared and it was concluded that most land use
characteristics are similar in both systems, especially for retail, 1. calculate generated trips based on linear regression
and both systems could be used to calculate the number of equations or average rates according to the ITE trip generation
generated trips. manual;
C. ITE Trip Generation Information System 2. calculate the number of internal trips in the research area;
Trip generation information system is developed by the 3. calculate the number of public transport trips;
Institute of Transportation Engineers [18] and provides 4. calculate the number of hiking and cycling trips.
regression equations and average transport rates to estimate
transport generated trips. The main disadvantage of ITE trip 6000
generation manual is the necessity of additional trip generation
5000
T=AverageVehicleTripEnds

result corrections for adjacent sites and public transport

availability near the site. 4000

3000
ITE trip generation manual regression equations for
2000
shopping centre are shown in Table I.
1000

TABLE I 0
100

120

140

160

180
0

20

40

60

80

ITE TRIP GENERATION MANUAL REGRESSION EQUATIONS FOR SHOPPING

X=1000sq.meters
CENTRE ITEFittedcurveresult RoDFittedcurve
Weekday morning peak hour 0.6LN(X) + 2.29
trips Y=e Fig. 3. Trip generation for shopping centre data
Weekday evening peak hour 0.66LN(X) + 3.40
trips Y=e A. Smart Growth Tools for ITE Trip Generation Information
0.65*LN(X) +5.83 System
Daily trips Y=e
Smart growth tools need to perform steps 24 mentioned
Weekend peak hour trips Y = e 0.65*LN(X)+3.78 above:
a) Mixed-use manual developed by ITE it reduces the
Average weekend trips Y = e 0.63*LN(X)+6.23 number of generated transport trips by reducing the number of
where x the shopping centre leasable area internal trips between different types of land use.
b) Urbemis (urban emissions) program [7] it estimates the
influence of sites on the environment (air quality, the
Despite the fact that some information systems take into
calculation of emissions) and evaluates the number of
account the availability of transport infrastructure near the site
transport generated trips for different land use types (3) and
and are used in different countries [2], it is necessary to take
takes into account information about public transport (4) and
into consideration:
pedestrian activity (5).
different transport trip rates between cities, regions,
countries;

90
 Information Technology and Management Science

___________________________________________________________________________________________________________ 2013 / 16
TABLE II
1.5h e
1 0.25 INITIAL DATA FOR TRIP GENERATION
1.5h e , (3)
Trip Reduction Gross
Actual generated trips,
0.0075 leasable
evening peak hours
Remarks
area, 2
where h = housing units, e = the number of employed Mixed use 1 32 000 1087
people in the research area. Improvements in transport
accessibility near the site
Mixed use 1 32 000 1462 (additional left turn, Fig.
Trip Reduction 0.075t + 0.075 p t , (4) 3)
Mixed use 2 22 000 1190
where t = the index of public transport services.
Mixed use 3 40 000 1593
The increase in gross
Mixed use 3 60 000 2025 leasable area
0.09 (d i + s c + b c )
Trip Reduction , (5)
3
site site
where: di the number of legs at intersections 2.5km2, sc
pedestrian way availability near the site, bc bicycle way
availability near the site.
c) Mix-used trip generation model ERA MXD it reduces the
number of generated trips taking into account the mixed land a) b)
Fig. 4. Transport accessibility: ) traffic organization scheme for mixed use 1;
use, number of employed people at the site and surrounding b) traffic organization scheme for mixed use 1
area (6).
B. Results
Trip Reduction T (1 - p i ) (1 - p w - p t ) , (6) Transport generated trips were calculated according to
TRICS, NZTPD and ITE trip generation information systems
where T the number of trips calculated by ITE trip (Table 3). Additional corrections for smart growth based on
generation information system, pi the probability of internal Urmebis, ERA MXD and NCHRP 851 tools were considered
trips in the research area, pw the probability of external for ITE trip generation information system. Per cent error as a
hiking or cycling trips, pt the probability of external public difference between observed and actual values for each
transport trips. information system was determined.
d) NCHRP 851 tool [8] it estimates the number of
internal transport trips between different types of land use and observed - actual
reduces the external transport generated trips by internal trip Error 100 (8)
actual
value (7).
Percent errors between observed and actual transport
d (b ) + d 2 (b 2 ) generated trips for evening peak hours for four mixed-use sites
Trip Reduction 1 1 , (7)
c are shown in Table III.

TABLE III
where b1 the space in building 1 of land use , b2 the
PERCENT ERROR BETWEEN OBSERVED AND ACTUAL TRANSPORT GENERATED
space in building 2 of land use , c the space in building 1 TRIPS FOR EVENING PEAK HOURS
and 2 of land use .
Observed ITE with correction, A/pm
Mixed trips, ITE ERA NCHRP TRICS NZTPD
IV. EXPERIMENTAL RESULTS A/pm Urmebis
use MXD 851
A. Initial Data 1 1087 34% 31% 17% 30% 10% 22%
Four sites with mixed land use and with gross leasable area 1 1462 0% 3% 38% 3% 18% 9%
from 29 to 60 thousand m2 were used in this study (Table II). 2 1190 5% 9% 28% 6% 31% 6%
The number of employed people outside the sites was
3 1593 6% 0% 13% 1% 6% 4%
calculated according to Riga City Development Plan for
3a 2025 10% 3% 13% 7% 10% 23%
20062018. Number of intersections, as well as percent of
streets with sidewalks was determined using publicly available
maps of the city. Number of public transport vehicles was where A/pm the number of transport generated trips for
calculated according to local municipality data; all evening peak hours.
The most accurate results were shown by TRICS
calculations were made for evening peak hours.
information system (Fig. 5, ) with 10% error and the ITE trip
generation information system with additional correction by

91
 Information Technology and Management Science

2013 / 16 ___________________________________________________________________________________________________________

ERA MXD smart growth tools with 17% error for mixed 1400
use 1. Increasing the transport accessibility (Fig. 5, b) near the 1200
site, changing traffic organization at adjacent intersection from

Generated trips, /peak hour

1000
mixed use 1, the best result was shown by the ITE trip 800
generation information system without any correction for 600
public transport and pedestrians trip generation result 400
accuracy was 99.6%. High calculation accuracy without any
200
additional corrections of the ITE trip generation information
0
system are related to the fact that mixed-use lands 1 are
ITE trip Urmebis ERA MXD NCHRP 8- Trics NZTPD
located near the freeway and till recently transport generation 51
accessibility from the site to the freeway was restricted information
system
(Fig. 4, ). After allowing the additional left turn from the site
to the freeway (Fig. 4, b), percent error significantly decreased Observed number of transport generated trips, evening peak hour

for the ITE trip generation information system without Actual number of transport generated trips, evening peak hour

corrections and for the ITE system with such smart growth Fig. 6. Transport generated trips. Mixed use 2
tools as Urbemis and NCHRP 8-51 was in the range of 0%
3%. If we take into account the accessibility of transport For mixed land use 2, the ITE trip generation information
infrastructure near the site (pedestrians, public transport, other system without additional corrections, ITE with NCHRP 8-51
sites near the mixed use 1), percent error for generated trips smart growth tool and NZTPD information system showed
grew and for such smart tools as Urbemis and NCHRP 8-51 the best results with percent error of 6% for generated trips
was 30%. (Fig. 6).
Mixed land use 3 and 3a provide an increase in the gross
leasable area without changing transport accessibility near the
1600
site. Accuracy of generated trip calculations for mixed land
1400
use 3 (Fig. 7, a) with the ITE trip generation information
Generated trips, /peak hour

1200
system and such smart growth tool as Urbemis was 99.9%.
1000
800
1800
600
1600
400
1400
Generated trips, /peak hour

200
1200
0 1000
ITE trip Urmebis ERA MXD NCHRP 8- Trics NZTPD 800
generation 51
information 600
system 400
200
Observed number of transport generated trips, evening peak hour 0
Actual number of transport generated trips, evening peak hour ITE trip Urmebis ERA MXD NCHRP 8- Trics NZTPD
generation 51
() information
system

Observed number of transport generated trips, evening peak hour

1600
Actual number of transport generated trips, evening peak hour
1400
Generated trips, /peak hour

1200 ()
1000
800
3000
600
2500
400
Generated trips, /peak hour

200 2000
0
1500
ITE trip Urmebis ERA MXD NCHRP 8- Trics NZTPD
generation 51 1000
information
system 500

Observed number of transport generated trips, evening peak hour 0

Actual number of transport generated trips, evening peak hour ITE trip Urmebis ERA MXD NCHRP 8- Trics NZTPD
generation 51
(b) information
system
Fig. 5. Transport generated trips: a) mixed use 1; b) mixed use 1a
Observed number of transport generated trips, evening peak hour
Actual number of transport generated trips, evening peak hour

(b)
Fig. 7. Transport generated trips: a) mixed use 3; b) mixed use 3a.

92
 Information Technology and Management Science

___________________________________________________________________________________________________________ 2013 / 16

High accuracy of calculations is related to mixed land Results of the considered mixed lands use showed that there
location outside the city centre with a good public transport was not the one best solution for transport generated trip
network and pedestrian walkways. Percent error for mixed calculations. The most reliable results were shown by the ITE
land use 3a increased in the range of 3%20% (Fig. 7, b) trip generation information system with additional correction
following an increase in the gross leasable area. by Urbemis with percent error up to 3%, for some mixed land
According to transport trip generation calculations, it could use accuracy was 99.9%. The complexity of smart growth
be seen that the presence or absence of public transport tool Urbemis application is that the number of reduced trips is
network and walkways near the site influenced accuracy in the shown in the daily section and not calculated for the peak
range of 3%10%. hours.
Trip generation results with the ITE trip generation Criteria of smart growth, such as the number of employed
information system without additional corrections, TRICS and people, public transport availability, walkability near the
NZTPD information systems did not show the one best research land use, and transport accessibility, could
method or tool for local conditions. For transport trip significantly improve or decrease results of calculations of
generation under local conditions, the most acceptable tool can generated trips for the transportation model, available initial
be considered the ITE trip generation information system with data, chosen linear regression equations and type of mixed
additional corrections by Urbemis smart growth tool. land use basis.
TRICS and NZTPD information systems showed
V. CONCLUSION insignificant accuracy of generated trip calculation; percent
Transportation trip generation models were evaluated taking error ranged from 6% to 30%. The reasons of such results
into account the mixed land use and transport infrastructure could be the small number of information system observations
availability in the research area. The number of transport for chosen conditions, incorrect adjustments for local
generated trips was calculated based on linear regression conditions and inaccuracy in input data.
equations. Accuracy of generated trips per hour was
maximized by the evaluation of smart growth criteria.
2006. [Online]. Available: http://en.scientificcommons.org/20188383i.
REFERENCES [Accessed May 31, 2013].
[1] J. D. Ortuzar and L. G. Willumsen. Modelling Transport, Fourth [15] Trips Rate Information Computer System. [Online]. Available:
Edition, 2011, John Wiley & Sons. http://www.trics.org/. [Accessed May 30, 2013].
[2] The Institute of Transportation Engineers, Trip Generation Handbook, [16] The Trips Database Bureau. [Online]. Available:
2nd ed. An ITE Recommended Practice, 2004. http://www.tdbonline.org. [Accessed May 30, 2013].
[3] J. Rhee, Improvement of Trip Generation Forecast with Category [17] Transport Roads nad Maritime Services. Road Projects. [Online].
Analysis in Seoul metropolitan area. Conference on the Eastern Asia Available: http://www.rta.nsw.gov.au/roadprojects. [Accessed May 30,
Society for Transportation Studies. 2003, Fukuoka, Japan. October 2003. 2013].
[4] J. Gonzalez-Feliu and F. Toilier and J.L. Routhier, End-consumer [18] Institutes of Transportation Engineers. [Online]. Available:
movements in French medium urban areas. The Sixth International http://www.ite.org. [Accessed May 30, 2013].
Conference on City Logistics, June 30 July 2, Puerto Vallarta, Mexico,
2009. [19] [13] R. Lee and J. Miller and R. Maiss and M. Campbell and K.
[5] I. Clark, Trip Rate and Parking Databases in New Zealand and Shafizadeh and D. Niemeier D and S. Handy, Evaluation of the
Australia. AITPM National Conference, National Convention Centre, Operation and Accuracy of Five Available Smart Growth Trip
Canberra, October 31- November 2, 2007. Generation Methodologies. [Online]. Available:
[6] M. Vrtic, P. Frhlich, A simultaneous disaggregate distribution and pubs.its.ucdavis.edu/download_pdf.php?id=1495. [Accessed May 30,
mode choice model: theory and application for a regional public 2013].
transport model. ERSA conference, July, 2005, Amsterdam,
Netherlands. Nadezda Zenina is a postgraduate student at the Faculty of Computer
[7] Nelson\Nygaard Consulting Associates, Adjusting Site-Level Vehicle Science, Riga Technical University (Latvia). She received her MSc. degree
Trip Generation Using URBEMIS, San Francisco, USA, Tech. Rep. from Riga Technical University, the Department of Modelling and Simulation
CA 94103, 2005. in 2006.
[8] B. S. Bochner and B. R.Sperry, Internal trip capture estimator for Since 2007 she has been working at Solvers Ltd (Latvia) as a
mixed-use developments, Tech. Rep. FHWA/TX-10/5-9032-01-1, Transportation and Modelling Engineer. Her skills cover the fields of
2007. transportation engineering, transportation planning and transportation
[9] A. Milne and S. Abley and M. Douglass, Comparisons of NZ and UK modelling. Research areas include artificial neural systems, data mining
Trips and Parking Rates, Tech. Rep. NZ Transport Agency Research methods learning trees, multinomial logit and discriminant analysis, cluster
Report 374, 2009. analysis, classification tasks, traffic modelling, transportation sustainability.
[10] A. Khalik and M. Al-Taei Amal, Prediction Analysis of Trip E-mail: [email protected]
Production Using Cross-Classification Technique, Al-Rafidain
Engineering, vol. 14, no. 4, pp. 51-63, 2006. Arkady Borisov received his Doctoral Degree in Technical Cybernetics from
[11] M. Mahmoud, M. Abdel-Aal, Cross classification trip production Riga Polytechnic Institute in 1970 and Dr.habil.sc.comp. degree in Technical
model for the city of Alexandria, Alexandria Engineering, vol. 43, no. Cybernetics from Taganrog State Radio Engineering University in 1986.
2, pp. 177-189, 2004. He is a Professor of Computer Science at the Faculty of Computer Science,
[12] R. Cervero and K. Kockelman, Travel Demand and the 3Ds: Density, Riga Technical University (Latvia). The research areas include artificial
Diversity, and Design, Transportation Research: D., vol. 2, no. 3, pp. intelligence, decision support systems, fuzzy set theory and its applications
199-219, 1997. and artificial neural systems. He has 235 publications in the field.
[13] R. Ewing and R. Cervero, Travel and the Built Environment: A Meta- He is a member of IFSA European Fuzzy System Working Group, Russian
Analysis, American Planning Association, vol. 76, no. 3, pp. 265-294, Fuzzy System and Soft Computing Association, honorary member of the
2010. Scientific Board, member of the Scientific Advisory Board of the Fuzzy
[14] A. Justen and U. Beuck, K. Nagel, Application of the VISEVA demand Initiative Nordrhein-Westfalen (Dortmund, Germany). E-mail:
generation software to Berlin using publicly available behavioral data, [email protected]

93
 Information Technology and Management Science

2013 / 16 ___________________________________________________________________________________________________________
Nadeda Zeina, rkdijs Borisovs. Regresijas analze transporta braucienu skaita novrtanai
Transporta attstbas prvadjuma modei pirmais solis klasiskaj etrsou transporta model: enerto transporta braucienu aprins, transporta sadaljums,
transportldzeku tipa izvle (piem., privtais un sabiedriskais transports) un transporta norkojums saska ar izvlto transporta tipu. Rakst ir izptti transporta
attstbas prvadjumu modei, lai uzlabotu enerts transporta plsmas aprina precizitti, emot vr jaukto apbvi un transporta infrastruktras
(sabiedriskais transports, gjji) pieejambu. Veikts informcijas sistmu apskats (Trics, ITE, NZTPD) un izskatti smart growth (Urmebis, ERA MXD,
NCHRP 8-514) instrumenti, kuri auj uzlabot enerts transporta plsmas aplses apjomus tuvk novrojamiem. Stratijas smart growth mris mazint
transporta braucienu skaitu pc principa augstie apbves blvumi kop ar jaukts zemes izmantoanu. Analze veikta taktiskaj lmen. Nodemonstrti enertas
transporta plsmas eksperimentu rezultti uz lineras regresijas viendojumu un smart growth instrumentu pamata. Analzes rezultti uzrdja, ka nav viena
vienga labka risinjuma no izskattajiem enerts transporta plsmas novrtjumam. Rezulttu precizitti ietekmja ne tikai dati par iedzvotju nodarbintbu,
sabiedrisk transporta esambu, bet ar ierobeojumi transporta infrastruktr un infrastruktras pieejambas paaugstinana (jauni atauti kreisie manevri
transportldzeku kustbai). Informcijas sistmas TRICS un NZTPD uzrdja zemu enertas transporta plsmas precizitti, kda starp aprina un novrojumo
transporta plsmu sastdja 6% - 30%. Viens no iemesliem zemajai precizittei ir neliels datu izlases apjoms. Analzes rezultti uzrdja, ka slikti attstts
sabiedrisk transporta un gjju tkls izptes objekta tuvum var samazint enertas transporta plsmas precizitti par 3% -10%. Pielietojot ITE metodi kop ar
Urbemis instrumentu, transporta enerto braucienu skaita aprina precizitte sastdja 99.9% (atsevim bvm). Augsta precizitte ir saistta ar izptes
objekta izvietojumu pilst pilstas mal ar labi attsttu sabiedrisk transporta un gjju tklu.

, .
- : ,
, .

( , ).
(Trics, ITE, NZTPD) smart growth (Urmebis, ERA MXD, NCHRP 8-514),
. smart growth
- .
.
smart growth. ,
. ,
, / . TRICS NZTPD
, 6% - 30%.
(
, 20-40 2). ,
3% - 10%.
Urmebis (
99,9%).

94