Is Random Regret Minimization More Suitable in Predicting Mode

Choice Decision for Indonesian Context than Random Utility

Maximization?

Muhammad Zudhy Irawan

, Sigit Priyanto

and Dewanti

Department of Civil and Environmental Engineering, Universitas Gadjah Mada, Yogyakarta, Indonesia

Keywords:

Travel Mode Choice, Multinomial Logit Model, Stated Preference Survey, Elasticity, The Value of Travel

Time Saving.

Abstract:

Since often encountered the missing prediction by using the concept of random utility maximization (RUM)

for Indonesian context, this study proposed a theory of random regret minimization (RRM) aiming to more

precisely predict the chosen mode and to increase the model ﬁt. Three variances of RRM were implemented:

Classical RRM, µRRM, and PRRM. Yogyakarta and Palembang were chosen as a case of the study by involv-

ing 708 respondents. A stated preference survey was carried out by offering six scenarios to the respondents.

We apply the value of ﬁnal log-likelihood, rho-square, Akaike and Bayesian Information Criterion, and hit rate

to compare the model ﬁt. We also calculate the value of travel time saving, and the time and cost elasticity. The

result shows that by excluding the rho square, RRM outperforms RUM in both cities. The µRRM produces the

best model ﬁt in a case of travel mode choice in Yogyakarta, while there is a tendency that PRRM produces

a better model ﬁt than µRRM in Palembang. We also found that RRM tends to generate a higher VTSS, time

and cost elasticity than RUM. Travellers in both cities also tend to be more sensitive to change in travel time

than travel cost.

1 INTRODUCTION

To date, numerous studies worldwide concerning

the choice decision use random utility maximization

(RUM)-based discrete choice model in predicting the

choice of several offered alternatives. This modelling

approach assumes that people choose one of sev-

eral options which have the highest utility (McFad-

den et al., 1973). In transportation studies, a logit

model is the most widely used method in the dis-

crete choice model (Ding et al., 2017) (Dong et al.,

2018). The RUM based discrete choice model also

applied in many studies of travel mode choice in In-

donesia. (Irawan et al., 2017) applied RUM-based

binary logit model to analyze the potential demand

of bus mode for egress trip from railway stations in

Yogyakarta, which is motivated by a situation that

train passengers prefer to opt to park their owned

motorcycle at destination railway station compared

to they have to use bus mode for their egress trip.

(Bastarianto et al., 2019) used RUM-based multino-

mial logit model, nested logit model, and cross-nested

logit model in understanding the joint choice of travel

modes and tour types for commuters from Bekasi

to Jakarta, Indonesia. (Irawan et al., 2018) applied

RUM-based ordered logit model in predicting the de-

mand of hybrid car in Indonesia. Meanwhile, (Rezika

et al., 2018) used bivariate ordered probit model in

estimating the urban railway demand in Yogyakarta.

Regarding the choice between public transport

and motorcycle mode in Indonesia, we assume that

the RUM-based discrete choice model might not be

appropriate. This is due to many travellers prefer to

use motorcycle mode to avoid the intolerable service

of public transport (PT). It is evident that even though

the Indonesian government has reformed the public

transport service in some cities in Indonesia, private

vehicle users especially motorcyclists still reluctant to

shift to public transport mode (Ilahi et al., 2015). In

Yogyakarta, people who decide to use public transport

must depart much earlier to minimize the lateness at

the destination point, such as workplace and school

(Irawan and Sumi, 2011).

It also should be noted that even though mo-

torcycle give the high utility caused by itsﬂexibility

(Irawan and Sumi, 2012; Irawan, 2019), the motorcy-

cle is a transport means that is the most often involved

of a trafﬁc accident . (Soehodho, 2017) showed that

Irawan, M., Priyanto, S. and Dewanti, .

Is Random Regret Minimization More Suitable in Predicting Mode Choice Decision for Indonesian Context than Random Utility Maximization?.

DOI: 10.5220/0009880601930199

In Proceedings of the 2nd International Conference on Applied Science, Engineering and Social Sciences (ICASESS 2019), pages 193-199

ISBN: 978-989-758-452-7

193

motorcycle acts as an aggravating factor on the sever-

ity level of a trafﬁc accident. By considering the fac-

tor of trafﬁc death and injuries, the motorcycle could

not provide the highest utility compared to public

transport mode. However, Indonesian travellers are

also reluctant to use public transport caused by poor

service provided (Joewono et al., 2016).

As mentioned above, since we assume that RUM

might not be appropriate for public transport ver-

sus motorcyclist case, we consider utilizing an al-

ternative modelling approach to utility maximiza-

tion. This alternative modelling approach contrary

to RUM tries to minimize disutility. Recently, some

researches have also used the disutility minimization

model called Random Regret Minimization or RRM

in predicting mode choice decision (Chorus, 2010)

(BELGIAWAN et al., 2017). In RRM, travellers

choose a speciﬁc travel mode in an attempt to min-

imize a disutility obtained from the other alternative

travel modes (Chorus et al., 2008).

With the introduction of RRM, many studies have

proven that RRM outperforms RUM in terms of

model ﬁt, prediction accuracy as well as the value

of travel time savings (VTTS) and elasticities (Leong

and Hensher, 2015) (Hensher et al., 2013). For

the Indonesian context, previous research has com-

pared RUM and RRM for mode choice decision in

Bali (Belgiawan et al., 2017) and Jakarta (Belgiawan

et al., 2019). However, their studies do not speciﬁ-

cally study motorcycle and PT as alternative choices.

Therefore, this paper aims to compare the use of util-

ity maximization (RUM) and disutility minimization

(RRM) approach in mode choice decision for Indone-

sian context. The mode choice alternatives that we

discuss is between PT which includes Bus and Light

Rail Transit (LRT), and motorcycles.

2 LITERATURE REVIEW

RRM was ﬁrst introduced by (Chorus et al., 2008).

It assumes that a traveller chooses a speciﬁc mode

of transport to minimize anticipated regret. Since

then, there are some variants of RRM. The ﬁrst is

the Classical RRM which is an improvement of the

original RRM. By re-analysing ten datasets used to

compare RUM and Classical RRM, (van Cranenburgh

et al., 2015) proposed µRRM and Pure RRM (PRRM)

model to improve the model ﬁt of Classical RRM. Re-

cently, the use of RRM is not only for mode choice

decision but also for park-and-ride lot choice (Sharma

et al., 2019), route choice (Mai et al., 2017) (Li and

Huang, 2017), driver choice of crash avoidance ma-

neuvers (Kaplan and Prato, 2012), freight transport

(Boeri and Masiero, 2014), activity start time and du-

ration (Golshani et al., 2018), and automobile fuel

choice (Hensher et al., 2013). (Chorus et al., 2014)

showed that out of 43 empirical studies, 15 studies

found that RRM’s performance is better than RUM,

while 13 studies show that the model ﬁt differences

between RUM and RRM are generally small.

This study attempts to implement the various

kinds of RRM for Indonesia context. The ﬁrst study

was begun by comparing Classical RRM and RUM

model in term of travel mode choice decision (i.e.,

bus rapid transit, feeder bus, motorcycle, and car) in

Denpasar Greater Area, Bali (Belgiawan et al., 2017).

By considering the value of Akaike Criterion (AIC),

Bayesian Criterion (BIC), rho square, and ﬁnal log-

likelihood, the result shows that the RUM model out-

performs Classical RRM. However, both model result

in the low model ﬁt. We predict that the poor ser-

vice of public transport in Bali might result in the bias

data because of the difﬁculties experienced by the re-

spondents when facing the stated preference survey.

We also predict that comparing the mode choice be-

tween car and motorcycle regarding travel time and

trip cost becomes less precise because of the respon-

dents’ characteristics of socioeconomics inherently

more cause it.

To ﬁll the research gap of the previous studies, Yo-

gyakarta and Palembang were chosen as case stud-

ies because of the satisfying service of existing pub-

lic transport (Irawan et al., 2017) (Budi and ZUS-

MAN, 2015). Our respondents also focus on motor-

cycle users since we attempt to understand the main

reason of choosing motorcycle mode is more caused

by the utility offered by motorcycle (RUM model) or

unacceptable disutility when using bus mode (RRM

model). Because the RRM could be implemented

with a minimum of three alternative modes (Cho-

rus, 2010), we added a light rapid transit mode as a

choice instead of motorcycle and bus mode. In this

study, we also compare the value of travel time sav-

ing, travel cost and travel time elasticities between

RUM and RRM. Previous studies showed that the

value of RRM-based elasticity is higher than RUM

based elasticity (Belgiawan et al., 2017) (Thiene et al.,

2012).

3 THEORETICAL BACKGROUND

3.1 Regret Function

In the CRRM framework, (Chorus, 2010) deﬁned that

the regret associated with an alternative m for person

n is determined by:

ICASESS 2019 - International Conference on Applied Science, Engineering and Social Science

194

CRRM

= a

+ R

CRRM

+ ε

= a

z6=m

ln(1 + exp[β

qzn

− X

qmn

)]) + ε

(1)

Where R

is random regret for an alternative m

for person n, R

is systematic regret for alternative

m for person n, ε

is unobserved regret for alterna-

tive m for person n, a

is alternative speciﬁc con-

stant, β

is the estimated parameter associated with

the generic attribute X

, X

qzn

and X

g ft

are values as-

sociated with generic attribute Xq

for, respectively,

person n choosing alternative z and m.

Meanwhile, the formula to calculate the regret

function for µRRM introduced by (van Cranenburgh

et al., 2015) was modiﬁed by dividing the coefﬁcient

of β

with µ. Furthermore, they also found that the

formula for systematic regret of the P-RRM model is

as follows.

PRRM

= a

+ Σ

PRRM

qzmn

(2)

where:

PRRM

qzmn

= {

z6=m

max(0,X

qzn

− X

qmn

)i f β

≥ 0

z6=m

min(0,X

qzn

− X

qmn

)i f β

< 0

(3)

3.2 Probability Function

There is no difference in determining the probability

of utility function (RUM) and regret function (RRM).

The probability function of CRRM and PRRM is

written as:

CRRM−PRRM

= exp(−Rmn)/Σ

mεZz=1

exp(−R

)

(4)

Since the µRRM includes a scale parameter µ) as

an additional degree of freedom to allows the ﬂexibil-

ity of the regret function level attribute, the probabil-

ity function of µRRM is calculated by multiplying µ

with the regret value (R).

4 DATA DESCRIPTION

A household-based face-to-face interview survey was

carried out from March to April 2017 and from

June to July 2017 in Palembang and Yogyakarta re-

spectively. The selected respondents were travellers

whose origin and destination points are located within

a radius of 500 meters from the LRT station in Palem-

bang, while the respondents were randomly selected

in Yogyakarta. The location selection purpose in

Palembang was to reduce the bias data in the model

since we only considered the variable of travel time

and trip cost. We did not take into account several in-

ﬂuenced variables related to the access and egress trip.

We also deﬁned the selected respondents are travellers

who work and make a daily trip with a minimum dis-

tance of 5 km from home to work. It is due to the

consideration that LRT can be an opt of travel mode

choice and also exclude the motorcycle captive users

as the respondents.

There are 401 and 307 respondents involved in

Yogyakarta and Palembang respectively. The inter-

view survey was conducted in all existing and planned

train stations, and it was proportionally distributed

based on population in each sub-district where the

rail station is located. With the aim to obtain a bet-

ter validity level of data, the questionnaire form was

designed as simple as possible. We were hoping that

the respondents are able to completely answer all of

the questions asked by a surveyor within ten minutes

interval of time. The questionnaire form was cate-

gorized into two items. The ﬁrst is the characteris-

tic of respondents (gender, age, and income), and the

second is the stated preference (SP) survey. The SP

questionnaire can be seen in Figure 1.

5 RESULTS AND DISCUSSION

5.1 Estimation Result

The result of the RUM and RRM based MNL model

is presented in Table 1. We use PythonBiogeme (Bier-

laire, 2016) in estimating the value of coefﬁcient and

calculating its model ﬁt. The result shows that the

parameter of travel time and cost are 1% signiﬁcant

with a negative value (as expected) for RUM, CRRM,

and µRRM in both cities. However, in the case of Yo-

gyakarta, the value of µ is signiﬁcant at 10%. On the

other hand, PRRM shows the insigniﬁcant coefﬁcient

regarding travel time.

Figure 1: The questionnaire form of SP survey.

Is Random Regret Minimization More Suitable in Predicting Mode Choice Decision for Indonesian Context than Random Utility

Maximization?

195

Taking into account the model ﬁt, we present

the model ﬁt consists of Final Log-likelihood (Fi-

nal LL), Rho-square, Akaike Information Criterion

(AIC), Bayesian Information Criterion (BIC), and hit

rate. The result shows that µRRM produces the small-

est ﬁnal log likelihood in both cities. For Yogyakarta,

the value of AIC and BIC for µRRM also indicates

the best ﬁt compared to RUM and other RRMs. How-

ever, rho-square for RUM in Yogyakarta is better than

the all variances of RRM. Meanwhile, for Palembang,

PRRM is the best model ﬁt for the value of AIC, BIC,

and rho-square. Looking into the hit-rate model ﬁt,

the model result for Yogyakarta produces a similar

value of hit-rate by 38.53%, while µRRM and PRRM

show the highest hit-rate in Palembang by 41.04%.

Due to this, we found that RRM performs better

than RUM, in which in a case of travel mode choice

in Yogyakarta, µRRM is the best model among the

other RRM models. Meanwhile, it is not yet clearly

detected whether µRRM or PRRM producing the best

model ﬁt in Palembang, in which µRRM produces the

best model ﬁt of ﬁnal log-likelihood, whereas PRRM

produces the best model ﬁt AIC, BIC, and rho-square.

5.2 Value of Travel Time Saving

The value of travel time savings (VTTS) is used to

measures the willingness to pay for a traveller due to

a travel time reduction. For CRRM, the VTTS can be

measured by (Leong and Hensher, 2015):

V T T S

CRRM

= 60x

δR

CRRM

/δT T

δR

CRRM

/δTC

60x

z6=m

− β

T T

/(exp[−β

T T

(T T

− T T

)] + 1)

z6=m

− β

/(exp[−β

(TC

− TC

)] + 1)

(5)

Where TTzn and TCzn represent travel time and

travel cost of person n choosing mode z as the com-

petitor of mode m, respectively. Meanwhile, the

VTTS for the µRRM model is obtained by modifying

the coefﬁcient of β with β/µ in Eq. (5). Moreover,

the VTTS for PRRM can be calculated by (van Cra-

nenburgh et al., 2015).

V T T S

PRRM

= 60x

δR

PRRM

/δT T

δR

PRRM

/δTC

60x

−βT T Σ

z6=m

T T

<T T

−βTCΣ

z6=m

T T

<T T

(6)

Table 2 presents the value of travel time saving

for both RUM and RRM. However, in contrast to

RRM, it should be noted that the performance of the

other modes does not inﬂuence the VTTS of a speciﬁc

mode produced by RUM. In RRM, the VTTS mea-

sures will increase or decrease conditionally on both

the number of available alternatives in the choice set

and the changes in the inﬂuenced variables of chosen

alternative and nonchosen alternatives. From Table

2, it can be seen that in all variance of RRM in both

cities, the VTTS for LRT mode is the highest and the

VTTS for bus mode is the lowest. It means that trav-

ellers in Palembang and Yogyakarta are willing to pay

much more expensive when using LRT mode if there

is a reduction in the travel time unit. This condition

makes sense since the travellers believe that LRT is a

travel mode promising timeliness of travel. However,

the opposite situation occurs in bus mode representing

that travellers are not willing to pay more due to the

reduction of travel time. The VTSS for bus mode pro-

duced by µRRM and PRRM in both cities is approxi-

mately half of the VTSS for LRT mode. Looking into

a situation that the VTTS for motorcycle mode tends

to higher than bus mode in all variance of RRM, it

represents that it will be challenging to shift motor-

cyclists to use bus mode in their daily trip as it now

happens.

Figure 2: Estimation result.

Figure 3: Value of travel time saving (IDR per hour).

Comparing the VTTS produced by RUM and

RRM in a case of Yogyakarta, all variances of RRM

produces the higher VTTS than RUM except in bus

mode. Meanwhile, by excluding the bus mode in

Palembang, the VTTS for CRRM is higher than

RUM, and the VTTS for PRRM is lower than RUM.

However, since we have not the VTTS of each re-

spondent either from the questionnaire survey or sec-

ondary data, we cannot check what the best model be-

tween RUM and RRM which could precisely estimate

the VTTS is.

ICASESS 2019 - International Conference on Applied Science, Engineering and Social Science

196

5.3 Demand Elasticity

Elasticity is used to measure the percentage change of

probability value caused by the change of correlated

attributes. (Ben-Akiva et al., 1985) showed the equa-

tion used to calculate the direct elasticities of RUM

model is as follows.

RUM

mn.X

qmn

δP

δX

qmn

δX

qmn

= (1 − P

)β

qmn

(7)

Where E

RUM

mn.X

qmn

is RUM-based elasticity for trav-

eller n on mode m which is related to variable X

qmn

and β

are speciﬁc attribute x for traveller n by mode

m and estimated the parameter of attribute x. p

the probability of traveller n chooses mode m. For

RRM based elasticity value, the formula for PRRM

and µRRM is similar as follows (van Cranenburgh

et al., 2015).

PRRM−µRRM

mn.x

qmn



−

δR

PRRM−µRRM

δX

qmn

+ Σ

m∈Zz6=mz=1

δR

PRRM−µRRM

δX

qmn



(8)

Meanwhile, the equation to calculate the elasticity

for CRRM is as follow (Hensher et al., 2013).

CRRM

mn.X

qmn

(−

δR

CRRM

δX

qmn

+ Σ

m∈Zz6=mz=1

δR

CRRM

δX

qmn

)

qmn

(9)

Figure 4 presents the measurement of cost and

travel time elasticities. As we expected, the sign of all

the travel time and cost elasticities produced by RUM

and RRM in both cities are negative, means that a re-

duction of travel time and cost of an alternative mode

will increase the percentage of probability in choos-

ing of that alternative mode. However, it should be

noted that the value of elasticity of costs and travel

time cannot provide an idea of whether RUM is better

than RRM or vice versa. These ﬁndings will be more

useful related to policy implementation. For exam-

ple: as the policymakers, they hope that the resulted

elasticity value is large enough to ease them to make

decisions to increase the demand for public transport.

Looking into the cost elasticity, both RUM and

RRM produce the lowest cost elasticity for motorcy-

cle modes in both cities. It means that with the change

in travel costs, the motorcycle users will be the most

reluctant travellers to switch to bus and LRT modes.

For example: in the case of µRRM in Palembang City,

a 10% increase in out of pocket would only cause a

10% decrease in the probability of using a motorcy-

cle, while for bus and LRT modes could decrease the

probability of modal usage by 12% and 14% respec-

tively.

Meanwhile, the highest travel cost elasticity is for

LRT mode for the RUM, CRRM, and µRRM mod-

els, and the bus mode for the PRRM model showing

that those mentioned travel mode will be easy to leave

by its passengers if there is a slight increase in ticket

costs. From Table 3, it also can be found that peo-

ple living in Palembang are more elastic in changing

travel mode caused by a variable of the trip cost com-

pared to people living in Yogyakarta.

On the travel time elasticity, the value produced

by the variances of RRM is not consistently higher

than RUM. Even though µRRM results in the high-

est travel time elasticity in Yogyakarta, both RUM

and CRRM produces a higher value than µRRM and

PRRM in Palembang. Different from the cost elas-

ticity, bus mode has the lowest travel time elasticity

in both cities meaning that with the change in travel

time, bus users are the most resistant travellers to use

the current mode. It is reasonable because people use

motorcycle or LRT mode is more caused by travel

time saving so that if there is a small increase of travel

time, motorcyclists and LRT passengers are the most

vulnerable travellers from the additional travel time.

Similar to the previous ﬁnding in cost elasticity, the

change in travel time is felt more signiﬁcant for peo-

ple living in Palembang than people in Yogyakarta.

Comparing between the elasticity of cost and

travel time, the travel time elasticity generated by

RUM and all variances of RRM is higher than the

cost elasticity, except for motorcycle and bus mode

in Yogyakarta produced by the µRRM model. This

situation represents that the change in the travel time

factor makes the traveller more sensitive to switch to

other modes compared to the shift in travel cost that

must be spent. Meanwhile, in a case where the elas-

ticity of cost is higher than the travel time, the authors

cannot ﬁnd the reason why did it happen. Therefore,

in further research, a more in-depth analyzis is needed

to reveal the phenomena that occur.

Finally, comparing among the elasticity values

produced by the RUM and RRM model, all variances

of RRM produces a higher elasticity than RUM. In

a more speciﬁc case, µRRM and CRRM result in the

highest elasticity value in a case of travel mode choice

in Yogyakarta and Palembang respectively.

Figure 4: Value of travel time saving (IDR per hour).

Is Random Regret Minimization More Suitable in Predicting Mode Choice Decision for Indonesian Context than Random Utility

Maximization?

197

6 CONCLUSIONS

This study implements all variances of RRM consist-

ing of CRRM, µRRM, and PRRM. To compare which

results are better between RUM and RRM, we use

the statistic tests of Final Loglikelihood, Rho-square,

Akaike and Bayesian Information Criterion, and Hit

Rate. Our modelling result indicates that RRM out-

performs RUM. Even though RUM still produce the

better rho square, RRM could produce the lower of

Final Loglikelihood and Akaike and Bayesian Infor-

mation Criterion. The probability of choice gener-

ated by RRM could estimate more precisely shown

by the value of hit rate and the average probability

value for chosen mode and non-chosen mode. Among

all variances of RRM, it can be generally concluded

that µRRM could produce the best model ﬁt although

there is a propensity that PRRM delivers a better

model ﬁt than µRRM in a case of travel mode choice

in Palembang.

The value of travel time saving produced by RUM

and RRM shows that RRM tends to provide a higher

VTTS than RUM. The highest VTTS in both cities

generated by RRM is on LRT mode indicating that

people are willing to pay more when using LRT if

there is a reduction in the travel time unit. Meanwhile,

the demand elasticity shows that the travel time elas-

ticity generated by RUM and all variances of RRM

(except µRRM in Yogyakarta) is higher than the cost

elasticity showing that travellers are more concerned

with the travel time than travel cost in deciding what

kind of transport means that they use. Both RUM

and RRM produce the lowest cost elasticity for mo-

torcycle mode and lowest travel time elasticity for bus

mode, saying that motorcyclists and bus users are the

most unwilling travellers to shift to other modes due

to the change of travel cost and travel time respec-

tively.

REFERENCES

Bastarianto, F. F., Irawan, M. Z., Choudhury, C., Palma, D.,

and Muthohar, I. (2019). A tour-based mode choice

model for commuters in indonesia. Sustainability,

11(3):788.

BELGIAWAN, P. F., Ilahi, A., and Axhausen, K. W. (2017).

Bali trans sarbagita: Comparison between utility max-

imization and regret minimization. In Proceedings of

the Eastern Asia Society for Transportation Studies,

volume 11. Eastern Asia Society for Transportation

Studies.

Belgiawan, P. F., Ilahi, A., and Axhausen, K. W. (2019). In-

ﬂuence of pricing on mode choice decision in jakarta:

A random regret minimization model. Case Studies

on Transport Policy, 7(1):87–95.

Belgiawan, P. F., Schmid, B., Dubernet, I., and Axhausen,

K. W. (2017). Comparison between rum, rrm variants,

and ram: Swiss sp and rp data sets. In 17th Swiss

Transport Research Conference (STRC 2017). STRC.

Ben-Akiva, M. E., Lerman, S. R., and Lerman, S. R. (1985).

Discrete choice analysis: theory and application to

travel demand, volume 9. MIT press.

Bierlaire, M. (2016). Pythonbiogeme: a short introduction

report transp-or 160706 transport and mobility labo-

ratory, school of architecture, civil and environmental

engineering, ecole polytecnique f

erale de lausanne.

Boeri, M. and Masiero, L. (2014). Regret minimisation

and utility maximisation in a freight transport context.

Transportmetrica A: Transport Science, 10(6):548–

560.

Budi, N. S. and ZUSMAN, E. (2015). Governing public

transport improvement program in indonesia:-a com-

parative case study of trans musi palembang and trans

metro bandung. Journal of the Eastern Asia Society

for Transportation Studies, 11:2528–2542.

Chorus, C., van Cranenburgh, S., and Dekker, T. (2014).

Random regret minimization for consumer choice

modeling: Assessment of empirical evidence. Jour-

nal of Business Research, 67(11):2428–2436.

Chorus, C. G. (2010). A new model of random regret min-

imization. European Journal of Transport and Infras-

tructure Research, 10(2).

Chorus, C. G., Arentze, T. A., and Timmermans, H. J.

(2008). A random regret-minimization model of travel

choice. Transportation Research Part B: Methodolog-

ical, 42(1):1–18.

Ding, C., Chen, Y., Duan, J., Lu, Y., and Cui, J. (2017). Ex-

ploring the inﬂuence of attitudes to walking and cy-

cling on commute mode choice using a hybrid choice

model. Journal of advanced transportation, 2017.

Dong, B., Ma, X., Chen, F., and Chen, S. (2018). Investigat-

ing the differences of single-vehicle and multivehicle

accident probability using mixed logit model. Journal

of Advanced Transportation, 2018.

Golshani, N., Shabanpour, R., Auld, J., and Moham-

madian, A. (2018). Activity start time and dura-

tion: incorporating regret theory into joint discrete–

continuous models. Transportmetrica A: transport

science, 14(9):809–827.

Hensher, D. A., Greene, W. H., and Chorus, C. G. (2013).

Random regret minimization or random utility max-

imization: an exploratory analysis in the context of

automobile fuel choice. Journal of Advanced Trans-

portation, 47(7):667–678.

Ilahi, A., WARO, A. I., and Sumarsono, P. (2015). Public

transport reform in indonesian cities. In Proceedings

of the Eastern Asia Society for transportation stud-

ies, volume 10. Eastern Asia Society for Transporta-

tion Studies.

Irawan, M. Z., Belgiawan, P. F., Widyaparaga, A., Budiman,

A., Muthohar, I., Sopha, B. M., et al. (2018). A mar-

ket share analysis for hybrid cars in indonesia. Case

studies on transport policy, 6(3):336–341.

Irawan, M. Z., Putri, M. K., Belgiawan, P. F., and Dwitasari,

R. (2017). Analyzing commuters’ behavior on egress

ICASESS 2019 - International Conference on Applied Science, Engineering and Social Science

198

trip from railway stations in yogyakarta, indonesia.

The Open Transportation Journal, 11(1).

Irawan, M. Z. and Sumi, T. (2011). Modeling departure

time of students on public transports in yogyakarta,

indonesia. Modern Applied Science, 5(4):3.

Joewono, T. B., Tarigan, A. K., and Susilo, Y. O. (2016).

Road-based public transportation in urban areas of in-

donesia: What policies do users expect to improve the

service quality? Transport policy, 49:114–124.

Kaplan, S. and Prato, C. G. (2012). The application

of the random regret minimization model to drivers’

choice of crash avoidance maneuvers. Transportation

research part F: trafﬁc psychology and behaviour,

15(6):699–709.

Leong, W. and Hensher, D. A. (2015). Contrasts of relative

advantage maximisation with random utility maximi-

sation and regret minimisation. Journal of Transport

Economics and Policy (JTEP), 49(1):167–186.

Li, M. and Huang, H.-J. (2017). A regret theory-based

route choice model. Transportmetrica A: Transport

Science, 13(3):250–272.

Mai, T., Bastin, F., and Frejinger, E. (2017). On the similar-

ities between random regret minimization and mother

logit: The case of recursive route choice models. Jour-

nal of choice modelling, 23:21–33.

McFadden, D. et al. (1973). Conditional logit analysis of

qualitative choice behavior.

Rezika, W. Y., Priyanto, S., and Irawan, M. Z. (2018). Esti-

mation urban railway demand in yogyakarta using bi-

variate ordered probit model. In MATEC Web of Con-

ferences, volume 181, page 02009. EDP Sciences.

Sharma, B., Hickman, M., and Nassir, N. (2019). Park-and-

ride lot choice model using random utility maximiza-

tion and random regret minimization. Transportation,

46(1):217–232.

Soehodho, S. (2017). Public transportation development

and trafﬁc accident prevention in indonesia. IATSS

research, 40(2):76–80.

Thiene, M., Boeri, M., and Chorus, C. G. (2012). Ran-

dom regret minimization: exploration of a new choice

model for environmental and resource economics. En-

vironmental and resource economics, 51(3):413–429.

van Cranenburgh, S., Guevara, C. A., and Chorus, C. G.

(2015). New insights on random regret minimization

models. Transportation Research Part A: Policy and

Practice, 74:91–109.

Is Random Regret Minimization More Suitable in Predicting Mode Choice Decision for Indonesian Context than Random Utility

Maximization?

199