See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/353163271

Mode Choice Analysis of Trips of Students and Staff in Education Hub Using
Multinomial Logit Model

Research · July 2021

DOI: 10.37896/jxu14.6/075

CITATION READS

1 832

2 authors:

Rena Shukla Himanshi Y Gadani

L. D. College of Engineering L. D. College of Engineering
19 PUBLICATIONS 12 CITATIONS 3 PUBLICATIONS 1 CITATION

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Rena Shukla on 11 July 2021.

The user has requested enhancement of the downloaded file.

Journal of Xidian University https://doi.org/10.37896/jxu14.6/075 ISSN No:1001-2400

Mode Choice Analysis of Trips of Students and Staff in

Education Hub Using Multinomial Logit Model
Rena N. Shukla #1, Himanshi Y. Gadani *2
#1
Civil Engineering Department, L.D. College of Engineering, India.
*2
Sardar Sarovar Narmada Nigam Limited, Government of Gujarat, India

Abstract— Transportation is one of the most important aspects of the overall development of any country, giving due importance to
transportation planning. In order to develop the travel demand model for a region, all type of sub-populations must be appropriately
represented; and students form a crucial fragment of population, whose travel behavior is quite unique and flexible making them
open to various alternative travel modes. The planning and management of any transportation system requires a detailed analysis of
the peoples’ mode choice behavior. It allows us to ascertain the preferences and public demands for different modes of transportation.
It also enables us to determine the transportation mode that is most probable. This paper studies the mode choice behavior of
students studying at various educational institutes located in the Navrangpura area of Ahmedabad. Navrangpura is an educational
hub and strongly representing education landuse in addition to residential & commercial landuse. The study collects the travel
information of the staff and students of the concerned colleges and universities with the help of an internet based survey. The primary
data analysis indicate the use of a variety of travel modes. The data was fed into TransCAD software to develop a Multinomial logit
model (MNL). The MNL model shows the relationship between various parameters like travel time (TT), travel distance (TD), travel
cost (TC) and income and its impact on different modes of transportation. This paper focuses on developing MNL model and deriving
the utility functions of different modes of travel. The utility model depends on various parameters – TT, TC, TD, socio-economic
status, convenience and availability of public transportation and is affected by numerous factors such as the change in price of the
fuel, changes in bus fares, parking availability etc.

Keywords— Mode choice, Multinomial Logit Model, TransCAD, Travel Cost, Travel Distance, Travel Time, Utility function.

I. INTRODUCTION
Mode choice analysis is an important step of four stage modeling in Transportation planning. The user
is choosing mode of travel from private vehicle, Intermediate public transportation vehicle and public
transportation vehicle as per socio economic background of the family. Also, the choice is not same
forever; it changes with time and age.
Students exhibit a unique travel behavior and are more flexible to try out different travel modes.
Moreover, their travel patterns indicate a certain amount of ambiguity; separating them from the general
population. A study based on the students’ travel behavior is therefore necessary in order to develop an
accurate travel demand model for the region. This becomes even more important for the areas having a
large university or a number of colleges /institutions, as a significant number of students and employees
would be travelling during a particular time of the day, causing congestion in the vicinity. Youngsters &
health conscious people cycle and trip by walk in society. Students also prefer to go by walk and by cycle
if living in vicinity of educational institute. Due to lack of income some students cannot afford and use
bicycle or go by walk for study trip. Students also prefer to use public transportation if concession and
monthly pass system is facilitated.
The mode choice study of the students indicates a share of different modes and thus all the modes of
transport are well represented in the analysis. The reason for a particular mode choice varies with the
individual and depends on a number of factors:
 Socio-economic factors – Household income, Car ownership, Family size and composition. Trip
Characteristics – Travel distance, Purpose of the trip, Time of the day,
 Availability of friend with vehicle to pick up and/or drop

VOLUME 14, ISSUE 6, 2020 661 http://xadzkjdx.cn/

Journal of Xidian University https://doi.org/10.37896/jxu14.6/075 ISSN No:1001-2400

 Pocket money given by parents and hidden expenditures in student life

 Characteristics of the transportation system – Travel time, Travel cost, Reliability factor, Comfort
and convenience.
The mode choice behavior can be better analyzed with the help of computational models. Mode choice
models analyze and predict the choice that an individual makes in selecting the transportation mode that is
used for particular type of trip. Characteristically, the goal of mode choice models is to forecast the
number of trips made by any particular mode. Also, one of the important objective in mode choice
modeling is to envisage the number of trips involving public transportation. The earlier mode choice
models were classified into two types: trip end model and trip interchange model. These mode choice
models, however, have become obsolete. Nowadays, logit and probit models are used: Binary logit model,
Discriminant analysis, Probit analysis, Multinomial logit model, Nested logit model, etc.
The multinomial logit model is a method used to calculate the effect of explanatory variables on a range
of dependent variables. The MNL model assumes that the individuals have a latent choice / utilities for
different modes of transportation and tend to prefer the one with the highest utility. The model gives the
probability of each mode as a function of the utility of all alternatives. Thus the MNL states that the
probability of using a travel mode is the ratio of the utility function of that particular mode to the sum of
the utility of all the modes used in the study.
The utility indicates the ‘use’ of any mode of transportation. The utility of any mode depends on the
travel attributes and individual characteristics. A utility function can be developed based on these
parameters. The utility function thus helps us rank the alternatives and also identify the one with the
highest utility.

II. STUDY AREA

Ahmedabad, the largest city and economical capital of Gujarat, is well known for its educational
institutions. Navrangpura area of Ahmedabad incorporates many colleges/ universities offering education
in varied fields – science, commerce, engineering, architecture, arts, law, medical, management, computer

Fig. 1 Location of the selected colleges shown in TransCAD

application, etc. and is known as the education hub of the city. Thus the roads in the vicinity are usually
congested during morning and evening peak hours, owing to the significant number of students and

VOLUME 14, ISSUE 6, 2020 662 http://xadzkjdx.cn/

Journal of Xidian University https://doi.org/10.37896/jxu14.6/075 ISSN No:1001-2400

employees enrolled at various colleges/ universities – L.D. College of Engineering, Gujarat University,
Government Polytechnic for Girls, Government Polytechnic Ambawadi, L. M. College of Pharmacy,
CEPT University, M. G. Science Institute and Shri Sahjanand Arts and Commerce College, H.L.
Commerce College, Saint Xaviers, Som-Lalit Institute of Management, L.D. Arts College etc.

III. DATA COLLECTION AND ANALYSIS

An internet based survey was carried out. The questionnaire was designed to include various aspects of
travel information including:

1. Personal attributes – age, gender, income, possession of driving license, residence area
2. Mode specific factors – nearest bus-stop (accessibility), travel mode, comfort level, availability
of seat
3. Trip-specific factors – travel distance, travel time, travel cost, origin, destination, trip purpose

The choice was given amongst six modes –

1. 2W: two wheeler,

2. 4W: four wheeler,
3. 3W: Auto rickshaw Three wheeler (special/individual)
4. 3W: Auto rickshaw on sharing i.e.Shuttle
5. Public Transport: Bus – AMTS (Ahmedabad Municipal Transport Service) & BRTS (Bus
Rapid Transit System),
6. Non motorised ecofriendly mode: bicycle and walking.

Mode-specific questions were included for the respondents opting for 3W (sharing), AMTS and BRTS.
The questionnaire was floated via email and WhatsApp.
A sample size of 3405 has been acquired out of a total estimated Student and staff population of 22650,
i.e. 15%. Total 1756 forms/ samples have been obtained from L.D. College of Engineering, 271 from
CEPT University, 342 from Gujarat University, 214 from Government Polytechnic for Girls, 282 from
Government Polytechnic, 178 from L.M. College of Pharmacy, 162 from M.G. Science Institute and 200
from Shri Sahajanand Arts and Commerce College.
Discrepancies can be observed in the overall mode choice behavior of the students coming from
different colleges or universities owing to a number of factors. Also, the travel behavior of employees is
found to be different from the students. The teaching staff is found to be highly dependent on private
modes of transportation; whereas, the students were shown to exhibit a different behavior as explained
below.
L.D. College of Engineering, Gujarat University, L.M. College of Pharmacy, M.G. Science Institute
and Shri Sahajanand Arts and Commerce College have students from different socio-economic
backgrounds, thus showing a fair distribution of private and public transport; whereas CEPT University
charges higher fees and has students coming from higher income groups, making private modes pre-
dominant. On the other hand the polytechnic colleges have younger students who may not possess a
driving license, thereby making them dependent on public transportation. The detailed mode choice
distribution is presented in the following table.

VOLUME 14, ISSUE 6, 2020 663 http://xadzkjdx.cn/

Journal of Xidian University https://doi.org/10.37896/jxu14.6/075 ISSN No:1001-2400

TABLE I
MODE CHOICE DISTRIBUTION FOR THE STUDENTS AND EMPLOYEES OF SELECTED COLLEGES

Name of the College 2W 4W 3W 3W AMTS BRTS Walk Bicycle Total

(special) (sharing)
CEPT University 133 46 28 5 1 5 45 8 271
Government Polytechnic 63 15 1 3 60 60 78 2 282
Government Polytechnic for 36 2 0 0 86 56 34 0 214
Girls
Gujarat University 191 6 3 2 81 24 34 1 342
L. D. College of Engineering 540 61 23 45 286 287 500 14 1756
L.M. College of Pharmacy 91 10 2 2 13 37 21 2 178

M.G. Science Institute 68 6 2 0 38 19 26 3 162

Shri Sahajanand Arts and 69 0 6 2 54 60 9 0 200
Commerce College
Total 1191 146 65 59 619 548 747 30 3405

Fig. 2 Chart showing the percentage of data collected from each college

IV. MULTINOMIAL LOGIT MODEL

The multinomial logit model allows us to find out the probability of choosing travel mode K if the
disaggregate or the ratio of travelers in aggregate case will select mode K based on the following equation:

VOLUME 14, ISSUE 6, 2020 664 http://xadzkjdx.cn/

Journal of Xidian University https://doi.org/10.37896/jxu14.6/075 ISSN No:1001-2400

V. MNL MODEL DEVELOPMENT IN TRANSCAD

The data collected form the survey was fed into TransCAD software in order to develop a multinomial
logit model. The model so developed was based on three parameters – travel distance, travel time and
travel cost. In order to input the data in any software it has to be divided into categories based on the study
requirements and then coding needs to be done. Six travel modes are considered and the travel parameters
are categorized as given in the table below.

TABLE III
CODING AND CATEGORIES FOR MNL MODEL

Mode of travel Travel Distance Travel Time Travel Cost (in Rs.)
1 Two Wheelers 1 < 2 km 1 < 15 minutes 1 < 10
2 Four Wheelers 2 2-5 km 2 15-30 minutes 2 10-20
3 Bus 3 5-10 km 3 30-45 minutes 3 20-30
4 Three Wheelers 4 >10 km 4 45-60 minutes 4 30-40
5 Bicycle 5 > 60 minutes 5 40-60
6 Walk 6 > 60

Now, based on this MNL model, the utility functions of different modes of travel were derived and the
same are stated below.
Utility function for two wheelers:
U (2W) = 2.11(TD) – 1.75(TT) – 0.325(TC)

Where,
TD = Travel Distance
TT = Travel Time
TC = Travel Cost
Utility function for four wheelers:
U (4W) = - 1.967(TD) – 0.789(TT) + 1.969(TC)

Utility function for bus:

U (Bus) = - 0.79(TD) + 1.24(TT) + 0.184(TC)

Utility function for three wheelers:

U (3W) = - 1.82(TD) – 0.147(TT) + 1.32(TC)

Utility function for bicycle:

U (Bicycle) = - 1.65(TD) – 0.259(TT)

VOLUME 14, ISSUE 6, 2020 665 http://xadzkjdx.cn/

Journal of Xidian University https://doi.org/10.37896/jxu14.6/075 ISSN No:1001-2400

Utility function for walk:

U (Walk) = - 0.864(TD) + 0.662(TT)

The negative coefficient indicates the reduction in utility of a mode whereas the positive coefficient
indicates the increase in utility with respect to that particular parameter. However, some contradictions can
be seen in the utility functions of four wheelers and three wheelers. For non-motorized modes of
transportation, only two parameters – travel distance and travel time are considered.

VI. MODEL CALIBRATION

The multinomial logit model is calibrated by calculating the probability for each mode of transportation.
The probability of each mode of transportation lies between 0 and 1 (0 < p < 1); and the total probability
P=1.
The utility equations with the parameters – travel distance = 1 (up to 2 km), travel time = 1(up to 15
minutes) and travel cost = 1(up to 10 Rs.), is given below:

U (2W) = 2.11 (1) – 1.75(1) – 0.325(1) = 0.041

U (4W) = - 1.967(1) – 0.789(1) + 1.969(1) = -0.787

U (Bus) = - 0.79(1) + 1.24(1) + 0.184(1) = 0.634

U (3W) = - 1.82(1) – 0.147(1) + 1.32(1) = -0.647

U (Bicycle) = - 1.65(1) – 0.259(1) = -1.916

U (Walk) = - 0.864(1) + 0.662(1) = -0.202

Now, the probability of each mode:

Probability p (2W) = eU(2W)

eU(2W) + eU(4W) + eU(Bus) + eU(3W) + eU(Bicycle) + eU(Walk)
= 0.213928265

P (2W) = 21.39%

Similarly,
p (4W) = 0.09347002 = 9.34%
p (bus) = 0.387083622 = 38.7%
p (3W) = 0.107516115 = 10.75%
p (bicycle) = 0.030224134 = 3%
p (walk) = 0.167777844 = 16.78%

Total Probability P = p(2W) + p(4W) + p(Bus) + p(3W) + p(Bicycle) + p(Walk)

= 1.

VOLUME 14, ISSUE 6, 2020 666 http://xadzkjdx.cn/

Journal of Xidian University https://doi.org/10.37896/jxu14.6/075 ISSN No:1001-2400

The probability for each mode of transportation changes continuously and depends on the relative
importance of the concerned parameters; nonetheless, the total probability is constant.

VII. CONCLUSIONS
Universities not just enroll a large number of students but is also a significant employer; attracting a
huge number of daily trips every day during morning and evening peak hours, making it an important
aspect of study in transportation planning. Eight colleges with a total population of 22,650 (students and
employees) were surveyed and a sample size of 3405 was obtained; i.e. 15%. The following conclusions
can be derived from the study.

 In case of students, a greater variation is noticed in the mode choice distribution. The number of
students using private vehicles, i.e. two wheelers and four wheelers, accounts to be one-third of
those using active modes of transportation, like walking, bicycle and public transportation. The
reason varies from non-possession of driving license to the socio-economic condition of their
families.
 The students living away from colleges and travelling a distance of more than 10 km, normally
prefer travelling in public transport bus (73%) rather than private vehicles (27%).
 All three parameter – travel distance, travel time and travel cost are found to be significant in the
mode choice of most students.
 The employees however, preferred private vehicles – cars and two wheelers over public
transportation, irrespective of travel distance, travel time and travel cost.
 The three wheeler (sharing) / jitneys were the dominant mode of transportation amongst the
housekeeping and supporting staff.
 Also the mode choice patterns of the students of the different colleges varied significantly. L.D.
College of Engineering, Gujarat University, L.M. College of Pharmacy, M.G. Science Institute and
Shri Sahajanand Arts and Commerce College have students from different socio-economic
backgrounds, thus showing a fair distribution of private and public transport; whereas CEPT
University charges higher fees and has students coming from higher income groups, making
private modes pre-dominant; public transportation accounted to only 3%. On the other hand the
polytechnic colleges have younger students many of who may not possess a driving license,
thereby making them dependent on public transportation; with just 15-20% students using private
vehicles.
 The utility of all transportation modes shows following results:
o The utility of two wheeler increases with the increase in distance, decreases with the
increase in travel time and also decreases with increase in travel cost; however travel cost
has a very low coefficient and so its effect can be considered negligible.
o The utility of bus is seen to increase with travel time and decrease with the travel distance
with negligible effect of travel cost. This is because bus is quite an economical mode of
transport and is mostly opted by the students coming from a distance or by those who do
not possess a driving license.
o The utility of four wheeler seems to increase with rise travel cost and decrease with the
increase in travel time and travel distance. The reason for the paradox may be that cars are

VOLUME 14, ISSUE 6, 2020 667 http://xadzkjdx.cn/

Journal of Xidian University https://doi.org/10.37896/jxu14.6/075 ISSN No:1001-2400

mostly used by the employees (professors) and students belonging to higher income groups
and they prefer four wheelers irrespective of any parameters.
o The utility of three wheelers decreases with the increase in travel distance and travel time,
as most would prefer other modes of transportation over longer distance and time. Though,
the utility increases with the increase in the travel cost; this is because among the
respondents, three wheelers were used by a very small group of people that did not have
access to private vehicles and did not prefer public transportation.
o The utility of bicycle tends to decrease with the increase in travel distance and travel time;
whereas for walking, the utility decreases with distance.

REFERENCES
[1] Das R., Kumar V., Prakash B., Dharmik & Subbarao S., “Analysis of University Students Travel Behaviour: En Route to Sustainable Campus.” Indian
Journal of Science and Technology, 2016.
[2] Hussain H.D., Mohammed A.M., Salman A.D., Rahmat R.A. & Borhan M.N., “Analysis of Transportation Mode Choice Using A Comparison of
Artificial Neural Network and Multinomial Logit Models”; ARPN Journal of Engineering and Applied Sciences, Vol. 12, No. 5, 2017.
[3] Khattak A., Wang X., Son S., & Agnello P., "Travel by University Students in Virginia: Is this Travel Different from Travel by the General Population?"
Transportation Research Record: Journal of the Transportation Research Board, 2011.
[4] Mazen D., Maya A.-Z., & Kaysi I., "Modeling Travel Choices of Students at a Private, Urban University: Insights and Policy Implications." Case Studies
on Transport Policy, Elsevier, 2014.
[5] Shukla R.N., Bhola A.D., Gundaliya P.J. & Zala L.B., “Mode Choice Analysis of Trips of Urban Residents using Multinomial Logit Model.” Journal of
Emerging Technologies and Innovative Research, Vol. 6, 2019.
[6] Whalen K. E., Paez A., & Carrasco J. A., "Mode Choice of University Students Commuting to School and the Role of Active Travel." Journal of
Transport Geography, Elsevier, 2013.
[7] Zhan G., Yan X., Zhu S., & Wang Y., "Using Hierarchical Tree-Based Regression Model to Examine University Student Travel Frequency and Mode
Choice Patterns in China." Transport Policy, Elsevier, 2016.
[8] Kadiyali L.R., “Traffic Engineering and Transportation Planning”; 8th Edn; Khanna Publishers, Delhi, 2013.
[9] Khisty C.J. & Lall B.K., “Transportation Engineering”; 3rd Edn; PHI Learning Private Limited, New Delhi, 2012.
[10] Sarkar P.K., Maitri V. & Joshi G.J., “Transportation Planning – Principles, Practices and Policies”; PHI Learning Private Limited, Delhi, 2017.

VOLUME 14, ISSUE 6, 2020 668 http://xadzkjdx.cn/

View publication stats