Computers & Industrial Engineering 59 (2010) 107–114

Contents lists available at ScienceDirect

Computers & Industrial Engineering

journal homepage: www.elsevier.com/locate/caie

Railway station site selection using analytical hierarchy process and data
envelopment analysis q
Nahid Mohajeri a, Gholam R. Amin b,*
a
Department of Architecture, Islamic Azad University, South Tehran Branch, Tehran 1584743311, Iran
b
Department of Computer Engineering, Postgraduate Engineering Centre, Islamic Azad University, South Tehran Branch, Tehran 1584743311, Iran

a r t i c l e i n f o a b s t r a c t

Article history: This paper deals with the problem of finding the optimum site for a railway station for the city of Mash-
Received 3 August 2009 had, northeast Iran, using the methods of analytical hierarchy process (AHP) and data envelopment anal-
Received in revised form 23 March 2010 ysis (DEA). The paper identifies a four-level hierarchy model for the railway station site-selection
Accepted 25 March 2010
problem. The model uses four main criteria: (1) rail-related, (2) passenger services, (3) architecture
Available online 27 March 2010
and urbanism, and (4) economics. In addition, there are 26 subcriteria as well as five (potential) candi-
dates or alternatives. Comparison matrices are used to obtain the local weights and priorities of the rail-
Keywords:
way-station candidates. A DEA model is proposed to determine the optimum site for a railway station. It
Railway station
Site selection
is shown that the local priorities (or weights) obtained from the AHP can be defined as the multiple out-
Multiple criteria puts of a DEA model for finding the best site for a railway station.
Analytical hierarchy process Ó 2010 Elsevier Ltd. All rights reserved.
Data envelopment analysis

1. Introduction and Alimohammadi (2009) suggested a fuzzy AHP method for

The aim of a site-selection problem is to find the optimum loca-
tion that satisfies a number of predetermined selection factors. The
process of a site selection typically involves two main stages:
screening and evaluating. The first stage identifies a limited number
of candidate sites, from a broad geographical area, taking into ac-
count the selection criteria. The second stage includes a careful
examination of alternatives to find the most appropriate site
(Chang, Parvathinathan, & Breeden, 2008). The second stage is an
important issue; for example, in the waste management where
the selection of an appropriate solid waste landfill site requires
the consideration of multiple alternatives and evaluation criteria
(Guiqin, Li, Guoxue, & Lijun, 2009). Choosing the location of a
new facility, such as a railway station, subject to multiple criteria
is an important decision-making problem for environmental man-
agers. In recent years, several decision-making methods have been
proposed for different site-selection applications. For example,
Ballis (2003) used the analytical hierarchy process (AHP) for an
airport-site selection on the Island of Samothraki, Greece. Also,
Guiqin et al. (2009) applied geographical information systems
(GIS) and AHP for solving the problem of selecting a landfill site
for solid waste in Beijing, China. Similarly, Vahidnia, Alesheikh,
q
This manuscript was processed by Area Editor Imed Kacem.
* Corresponding author. Tel./fax: +98 21 88347425.
E-mail address: g_amin@azad.ac.ir (G.R. Amin).
determining the optimum site for a hospital.
This paper explores the problem of finding the optimum loca-
tion of a railway station in the city of Mashhad, northeast Iran,
using a hierarchy structure. We introduce the railway station
site-selection problem as a hierarchy model consisting of four lev-
els, each with its own main criteria. The main criteria are: (1) rail-
related, (2) passenger services, (3) architecture and urbanism, and
(4) economics. Each of these main criteria is then divided into sev-
eral subcriteria, giving a total number of 26 subcriteria. In addition,
the hierarchy model has five potential railway stations, as candi-
dates or alternatives. We use expert judgment to perform the indi-
vidual pairwise comparisons in the AHP. Furthermore, we
incorporate a data envelopment analysis (DEA) for aggregation of
the AHP global priorities. The result indicates that the DEA model
is useful for finding global priorities among the potential railway
stations. In particular, we conclude that the local priorities (or
weights) obtained from the AHP can be defined as the multiple
outputs of a DEA model for finding the most suitable site for a
railway station.
2. Site selection using AHP
The AHP, initiated by Saaty (1980), is a flexible multicriteria
decision-making methodology that transforms a complex problem
into a hierarchy with respect to one or more criteria. The AHP
method has been used for a wide variety of decision makings in
fields such as government, business, industry, healthcare, and

0360-8352/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cie.2010.03.006
108 N. Mohajeri, G.R. Amin / Computers & Industrial Engineering 59 (2010) 107–114

education (Boroushaki & Malczewski, 2008; Forman & Gass, 2001; adjacent regions. Because of the position and recreational-religious
Jyrki et al., 2008; Linkov, Satterstrom, Steevens, Ferguson, & Pleus, role of city of Mashhad, there is a great number of activities in the
2007; Raharjo, Xie, & Brombacher 2009; Saaty, 2008), and also for city, many of which depend on the railway. In addition, the Mash-
site-selection problems. For example, Ballis (2003) used the AHP had railway station serves six main lines and eight secondary lines,
method for an airport-site selection on the Island of Samothraki, whose total length inside the station is about 28 km. The station is
Greece, and Korpela, Lehmusvaara, and Nisonen (2007) selected a equipped with four passenger platforms, but these are not suffi-
warehouse operator network using a combination of the AHP and cient to support the expected increase in the number of passengers
DEA methods. Also, Onut and Soner (2008) used the method for and associated rail services that are needed. The Mashhad railway
trans-shipment site selection and Rosenberg and Esnard (2008) station occupies an area of 175 hectares and is located in a high-
used a hybrid version for a transit site selection. Furthermore, density city centre. The railway lines thus conflict with the sur-
Hsu, Tsai, and Wu (2009) used the method to analyze tourist rounding urban structures. The Mashhad railway station needs to
choice of destination, Dagdeviren, Yavuz, and Kilinc (2009) to be improved for several reasons, including the following: (1) the
analyze the problem of weapon selection, and Garcia-Cascales limited capacity of the station building, (2) the small size of the
and Lamata (2009) to choose a cleaning system for engine mainte- railway fleet (rolling stocks), (3) the lack of a proper infrastructure,
nance. The AHP method requires the following pairwise compari- (4) the environmental problems such as visual and sound pollution
son matrix, A, which contains the relative weights of the criteria: around the station, and (5) an expected increase in the flow of pas-
0 w1 w1 w1 1 sengers and cargo transferred through the station in the near

w1 w2 wn future.
B C
B .. .. .. .. C There are several alternatives in dealing with the expected in-
B .
B . . . CC crease in the number of passengers and the volume of cargo. The
Bw
A ¼ B w1i
wi
w2
   wwni C
C first alternative is to develop an entirely new high-capacity station,
B C
B . .. .. .. C for which there are two candidates as potential sites. The second
B ..
@ . . . C
A alternative is to construct a satellite station adjacent to the present
wn wn wn
w1 w2
   wn station, for which there are also two candidates as potential sites.
The third alternative is to increase the capacity of the existing sta-
where wi is the importance weight of the ith criteria with respect to tion. Thus, the main purpose of this study is to find which of these
goal, or the importance weight of the ith subcriteria (i = 1, . . . , n) five options should be selected; that is the first alternative (with
with respect to criteria and so on. Furthermore, the importance two potential sites), the second alternative (again with two poten-
weights can be obtained using the following equation (Saaty, tial sites), or the third alternative.
1980, 2008).
Aw ¼ kmax w
3.1. A hierarchy model
where, kmax is the maximum eigenvalue of the matrix and
w = (w1, . . . , wn) is the corresponding eigenvector of A. Fig. 1 shows a four-level hierarchy model for the railway station
site-selection problem in Mashhad.
3. Information background The first level presents the goal of the problem, which is to find
the optimum place of the railway station among potential candi-
The Khorasan Razavi province is located in northeast Iran. The dates. As is shown in the second level, the objective of the model
Mashad railway station is a multi-functional facility located in is divided into four main criteria, namely rail-related, passenger
the centre of the city of Mashhad. This central region is in the vicin- services, architecture and urbanism, and economics. The third level
ity of the eastern and north-eastern railways. In addition, the train consists of 26 subcriteria which are related to the main criteria.
stations Neghab and Kashmar, both of which are close to the Also, the five potential stations are given at the final level of the
border between Iran and Turkmenistan, have end stations in the proposed hierarchical model.

Goal Criteria Subcriteria Alternative

C11
C12
C13
Rail related C14
C15
A 1: Move to Salaam
C16
C17
C21 A 2: Renovation Current station
C22
Passenger
C23
Services C24 A 3: Redevelopment of current
Railway Station
C25 and establishing a satellite station
Site Selection C26
C31
C32 A 4: Move to Paj village
Architecture C33
C34
and Urbanism C35
C36 A 5: Satellite station next to
C37 underground station
C38
C41
Economics C42
C43
C44
C45

Fig. 1. Hierarchy model for a railway station.

 N. Mohajeri, G.R. Amin / Computers & Industrial Engineering 59 (2010) 107–114
3.2. Choosing potential alternatives
Choosing potential alternatives for the railway station site, as
well as selecting the related main criteria and subcriteria, was
made using the results of transport and urban, rail-related, envi-
ronmental, and economic studies. In addition, the constraints re-
lated to the sites, world standards, regulations, and principles for
site selection were also considered (American Planning Associa-
tion., 2006; AREMA, 2006; Nijkamp, 2004; Ross, 2001). More spe-
cifically, the main principles are as follows:
 Travel time to railway station.
 Distance from railway station to public transport and other
facilities.
 Safety and coordination with civil-defence facilities.
 Compatibility and harmonization with the environment.
 Adjustment to the present and likely future city structures.
Using these main principles, four criteria, 26 subcriteria and 15
alternatives for site selection were studied. From these, the follow-
ing five alternatives were selected as having the highest potential:
A1: Move to the Salaam station.
A2: Renovation and development of the present railway station.
A3: Redevelopment of present station and establishing a satel-
lite station.
A4: Move to the village of Paj.
A5: Establish a satellite station next to one of the underground
stations.
3.3. The alternatives
Fig. 2 shows the place of the above-selected railway stations,
except to A3.
The first potential alternative, A1, is located outside of the boar-
der of the city, 18 km from the current station (Fig. 2). The second
Fig. 2. Locations of the alternatives.
109
site, which is the current railway station, A2, is located in the cen-
tral part of the city and occupies 175 hectares of valuable land. A4 is
a historical-recreational village and 18 km from the current station
and A5 is establishing satellite station, next to one of the under-
ground stations (Fig. 2). Also, the third potential site, A3, can be
viewed as a combination of two alternatives A2 and A5 and there-
fore is not shown in Fig. 2.
In the following we give a brief description for the five candi-
date sites:
A1: ‘‘Move to Salaam station”. This station is outside of the boar-
der of the city, 18 km far from the current station. Relocation of the
current railway station to this area would make the current site of
the train station available for new activities in accordance with the
comprehensive plan of the city of Mashhad. The valuable land that
would thus be free includes, in addition to the land occupied by the
station itself, the land occupied by the railway line, which has a
width of about 34 m and a total length of about 4000 m. Further-
more, this alternative provides a suitable access and communica-
tion to the neighbour rail stations. Further development of the
new station is another potential of this alternative. However, the
high cost of a new station and the long distance for daily trips of
the passengers from this station to the city centre are among draw-
back of this alternative.
A2: ‘‘Renovation and redevelopment of current railway station”.
This alternative aims at optimizing the capacity of present sta-
tion so as to cope with the predicted increase in passenger flow
and to provide suitable access to the city centre at the same
time. The suitable access to city centre and the small distance
from railway station to destination of the majority of Mashhad’s
passengers, namely the holy shrine, are the greatest advantages
of this alternative. In addition, the ease and quick access to all
public transportation and facilities for religious tourists are
important factors. This alternative, however, may cause some
environmental pollution to the surrounding residential areas,
such as visual, air and noise pollution, and possibly create an
unsuitable urban edge.
110 N. Mohajeri, G.R. Amin / Computers & Industrial Engineering 59 (2010) 107–114

A3: ‘‘Redevelopment of current station and establishing a satellite Table 1

station”. A combination of A2 and A5, and would be located 4 km The main criteria and subcriteria.

from current station and close to the border of the city. This alter- Main criteria Subcriteria
native provides the use of the current station only for electrical Rail-related C11 Feasibility of connecting new and recommended
rolling stock and building a new satellite station for diesel trains. lines to the station
This alternative makes the expensive land that is currently used C12 Possibility of creating and developing rail lines in
for the railway line free for other uses. The idea is that these tracks the
C13 Possibility of using the existing equipments
will be moved underground and used only for electrical rolling C14 Possibility of accessing the existing rail network
stock, thereby making the surface land free for other uses. The C15 Communication between adjacent stations with
satellite station is then supposed to be next to a planned under- regard to station services
ground station (at the location of A5 in Fig. 2). Because of the C16 Adapting to the geometric limitation of railway
line
combination with the current station and the underground, the
C17 Coordinating optimum movements and train
propose satellite station provides a minimum distance to general manoeuvres
destination points throughout the city of Mashhad.
Passenger C21 Possibility of offering suitable services to the
A4: ‘‘Move to village of Paj”. The village, located 18 km far from services passengers at the station
the current station, is expected to change from its current residen- C22 Possibility of most favorable access to the city
tial function into a historical-recreational area in the near future. transport systems
This new function potentially increases the number of passengers C23 Possibility of most favorable access to other
intercity transport systems
and offers possibilities so as to the further development of the sta-
C24 Proximity of the station to the destination of the
tion. In addition, it makes free the expensive land of the current majority of passengers
railway station and allows new land use at the site to increase its C25 Reduction in journey time
value. The great distance from this potential site to city centre C26 Easy access to urban facilities at the time of
accidents inside the station
for the passengers is one of the drawbacks of this alternative.
A5: ‘‘To establish a satellite station next to one of the underground Architecture and C31 Harmony with other adjacent utilities close to the
urbanism site
station”. Southeast of the city is an alternative that allows for con-
C32 Coordination with Mashhad urban development
siderable urban development in the vicinity of, and inside, the sta- plans
tion in the future. In addition, this option provides an easy access C33 Possibility of effective land possession
to the city centre through the underground station linked to the C34 Harmony with the image of city
railway station. While this location is far from the city centre it C35 Harmony with environment (Noise/air/visual
pollutions)
has positive aspects in that it decreases the risk of environmental
C36 Existence of potential for future development
pollution. C37 Easy access to the infrastructural installations
C38 Coordination with non-operative defense

3.4. Assessment criteria Economics C41 Potentials for return of investment cost
C42 Impact on opportunity cost of existing jobs
C43 Possibility of creating added value for the region
Based on the most important issues in the site selection of a C44 Decrease in passengers costs
railway station, we have selected the four main criteria as rail C45 Possibility of preserving existing useful land uses
and passenger services, urban planning and architecture, the envi-
ronment, and economics and related subcriteria, as are commonly
used in the literature (American Planning Association, 2006; ARE-
rail experts, five urban design and planning specialists, two envi-
MA, 2006; Nijkamp, 2004; Ross, 2001). Table 1 summarizes the
ronmental experts, four architecture and three economic special-
main criteria and the related subcriteria.
ists. Then we computed the weights of the corresponding
pairwise comparison matrices. Table 2 shows the importance
4. Determining the AHP weights weights obtained from the pairwise comparison of the criteria with
the goals.
Fig. 1 shows the proposed hierarchy model for the railway sta- Table 2 shows that the inconsistency is less than 0.1 and conse-
tion site-selection problem, where the overall objective is in the quently that the obtained priorities are reliable. Also, Tables 3a–d
first level. Also, the figure shows the main criteria in the second le- give the priorities of the subcriteria with respect to the main crite-
vel, 26 subcriteria in the third, and five potential stations in the last ria. The importance weights of the rail-related subcriteria, with re-
level. According to the AHP method, the elements of each level are spect to the rail-related criterion, are shown in the following table.
pairwise compared with the element in the next higher level. This The subcriterion C14, possibility of accessing the existing rail
results in a number of pairwise comparison matrices (Saaty, 1980, network, has the highest importance weight among the other
2008). We assess the importance of the ith criteria against the jth rail-related subcriteria (Table 3a). The priorities of subcriteria cor-
criteria using the following five-point assessment values. responding to the passenger services, architecture and urbanism
and economics are given in the following tables.
(i) If criteria i and j are equally important then the correspond-
ing element of the comparison matrix will be
w Table 2
aij ¼ w
wj
i
¼ 1&aji ¼ wj ¼ 1. Importance weights of criteria.
i
(ii) If criteria i be moderately more important than criteria j then
wi wj Goal Criteria Importance
aij ¼ w ¼ 3&aji ¼ w ¼ 13.
j i weight
(iii) If criteria i be extremely more important than criteria j then
w Railway station site Rail-related 0.238
aij ¼ w
w
i
¼ 5&aji ¼ wj ¼ 15.
j i selection Passenger services 0.172

Also, values 2 and 4 are used to show an intermediate impor- Inconsistency = 0.00092 Architecture and 0.358
urbanism
tance between the criteria. Using the hierarchy model and the cri- Economics 0.232
teria we developed standard questionnaires that were filled by five
 N. Mohajeri, G.R. Amin / Computers & Industrial Engineering 59 (2010) 107–114 111

To compute the priorities of five potential stations respect to Table 3d

the subcriteria, 26 pairwise matrices must be constructed. To re- Priorities of economic subcriteria.

duce the number of comparisons, we compare the alternatives Criteria Subcriteria Priorities weights
with respect to a cluster of rail-related subcriteria at a time, then Economics C41 0.410
with respect to a cluster of subcriteria related to the passenger ser- C42 0.087
vices, and so one. Tables 4a–d give the corresponding weights of C43 0.197
the alternatives respect to the criteria. Inconsistency = 0.00756 C44 0.162
The inconsistency values of all above comparisons are lower C45 0.144
than 0.1, indicating that the expert’s judgments are useful. The
AHP aggregated result is shown in Table 5.
The highlighted alternative has the highest importance weight Table 4a
among the five candidate stations and is the optimum place for Importance weights of alternatives.
the development of Mashhad railway station. The next section pro- Criteria Alternatives Overall
vides a DEA model for aggregation of the AHP priorities. weights
Rail-related A1: Move to Salaam 0.213
A2: Reorg. current 0.217
5. Aggregation using DEA Overall A3: Redev. of current and establishing a 0.218
inconsistency = 0.01 satellite station
The DEA is a linear programming method for measuring the rel- A4: Move to paj 0.121
A5: Satellite station next to underground 0.230
ative efficiency of decision making units (DMUs). The method was
initiated in operational research (OR) by Charnes, Cooper, and
Rhodes (1978) and developed by Banker, Charnes, and Cooper
(1984). In recent years there have been seen several successful Table 4b
applications of DEA models for evaluation of many different kinds Importance weights of alternatives.
of DMUs (Cooper, Seiford, & Tone 2007; Emrouznejad, Parker, & Criteria Alternatives Overall
Tavares 2008). For instance, Managi and Karemera (2004) devel- weights
oped a DEA technique for measuring productivity in the US agricul-
Passenger services A1: Move to Salaam 0.067
ture. Lindebo (2004) suggested a DEA method for the evaluation of A2: Reorg. current 0.278
the economic capacity of the Danish fishing fleet. Wang and Chin
Overall A3: Redev. of current and establishing a 0.331
(2007) introduced a DEA model for aggregation of a preference vot- inconsistency = 0.04 satellite station
ing system. Also, Amin and Emrouznejad (2010) originated the use A4: Move to paj 0.044
A5: Satellite station next to 0.279
underground
Table 3a
Priorities of rail-related subcriteria.

Criteria Subcriteria Priorities weights

Table 4c
Rail-related C11 0.15 Importance weights of alternatives.
C12 0.166
Criteria Alternatives Overall
C13 0.089
weights
C14 0.239
C15 0.101 Architecture and A1: Move to Salaam 0.150
urbanism A2: Reorg. current 0.187
Inconsistency = 0.00633 C16 0.175
C17 0.080 Overall A3: Redev. of current and establishing a 0.268
inconsistency = 0.02 satellite station
A4: Move to paj 0.134
A5: Satellite station next to 0.260
Table 3b
underground
Priorities of passenger services subcriteria.

Criteria Subcriteria Priorities weights

Passenger services C21 0.207 Table 4d
C22 0.198 Importance weights of alternatives.
C23 0.066
C24 0.124 Criteria Alternatives Overall
weights
Inconsistency = 0.02 C25 0.086
C26 0.319 Economics A1: Move to Salaam 0.108
A2: Reorg. current 0.245
Overall A3: Redev. of current and establishing a 0.278
Table 3c inconsistency = 0.01 satellite station
Priorities of architecture and urbanism subcriteria. A4: Move to paj 0.152
A5: Satellite station next to 0.217
Criteria Subcriteria Priorities weights
underground
Architecture and urbanism C31 0.264
C32 0.123
C33 0.130
C34 0.137 of a DEA model for finding the most relevant results of multiple
C35 0.083
search engines. In addition, DEA technique is used for obtaining
Inconsistency = 0.00815 C36 0.110 the weights or priorities of the comparison matrix in the AHP
C37 0.071
C38 0.081
method (Ramanathan, 2006; Wang & Chin, 2009; Wang, Liu, & El-
hag 2008).
112 N. Mohajeri, G.R. Amin / Computers & Industrial Engineering 59 (2010) 107–114

Table 5 Table 7
Importance weights of alternatives. Efficiency scores of DMUs.

Alternatives Aggregated weights Alternatives (DMUs) A1 A2 A3 A4 A5

A1: Move to Salaam 0.150 Efficiency 0.4913 0.8466 1 0.4119 0.9005
A2: Reorg. current 0.187
A3: Redev. of current and establishing a satellite station 0.268
A4: Move to paj 0.134
A5: Satellite station next to underground 0.260 where, e is a non-Archimedean and can be computed using the fol-
lowing model.
Here, we propose a new DEA method for aggregation of the AHP
result as follows. Let us assume we have n potential alternatives in emax ¼ max e
a hierarchy model. Also, suppose that we have m criteria. Let wij be s:t: 0:213u1 þ 0:067u2 þ 0:150u3 þ 0:108u4 6 1
the priority of the ith alternative with respect to the jth criteria ob- 0:217u1 þ 0:278u2 þ 0:187u3 þ 0:245u4 6 1
tained from the corresponding pairwise comparison in the AHP 0:218u1 þ 0:331u2 þ 0:268u3 þ 0:278u4 6 1 ð3Þ
method (i ¼ 1; . . . ; n; j ¼ 1; . . . ; m). To find the score of the kth po-
0:121u1 þ 0:044u2 þ 0:134u3 þ 0:152u4 6 1
tential alternative we propose the following DEA model (for each
k = 1, . . . , n). 0:230u1 þ 0:279u2 þ 0:260u3 þ 0:217u4 6 1
uj  e P 0 j ¼ 1; . . . ; 4
X
m
zk ¼ max wkj uj The optimal value of model (3) gives the maximum non-Archi-
j¼1
medean emax ¼ 0:9132 corresponding to the DMUs given in Table 6.
X
m
ð1Þ
s:t: wij uj 6 1 i ¼ 1; . . . ; n Now we solve model (2) for all the alternatives taking e ¼ emax ¼
j¼1 0:9132. Table 7 shows the efficiency scores of the alternatives.
uj P e j ¼ 1; . . . ; m
where zk is the score of the kth alternative (k = 1, . . . , n), and e is a 6. Results
non-Archimedean. Also, we define each potential alternative as a
DMU and the AHP priorities weights wij as the jth output of the Table 7 shows the results of the efficiency scores of the five
ith alternative (i ¼ 1; . . . ; n; j ¼ 1; . . . ; m). Note that in this applica- alternatives using the proposed DEA model. According to the
tion each DEA observation (or alternative) has only m outputs with scores of alternatives (or DMUs) the third unit, A3, is the only effi-
no input. Therefore, to compare the alternatives we assume that all cient DMU. Therefore, the third alternative is the most suitable for
the DMUs have the same amount of input, which is called a dummy Mashhad railway station site. According to the global priorities of
input (see Emrouznejad et al. (2008) for a full bibliography of appli- the alternatives obtained from AHP method, shown in Table 5,
cations). Now we apply the above model for the Mashhad railway the third alternative has the highest weight, 0.268. Also, the next
station site-selection problem. Tables 4a–4d give the priorities of preferred alternative is A5 with the weight of 0.26.
the five potential alternatives respect to the main criteria. Therefore The result obtained from the combined DEA and AHP method
the corresponding data for the proposed DEA model (1) is shown in also gives the same ranking. According to the efficiency scores gi-
Table 6. ven in Table 7, alternative A3 is the most preferred. This is followed
where DMUi represents the ith alternative (i = 1, . . . , 5), Oj by alternative A5, as the second best, and so on. The combined DEA
shows the jth output (j ¼ 1; 2; 3; 4), O1 are for the priorities of the and AHP aggregation method proposed in this paper takes the local
five alternatives with respect to rail-related, O2 for passenger ser- AHP priorities as multiple outputs. The combined DEA and AHP
vices, O3 for architecture and urbanism, and O4 for economics. Note method compares the potential railway stations and selects the
that Table 6 shows the required information for applying the pro- most highly preferred railway station. However, in this specific
posed DEA model (1) where the last column is used to show the application, the results of AHP combined with DEA are the same
single input of the alternatives. As the potential railway stations as when AHP is used alone. The primary reason that we combine
have multiple outputs (the local priorities) with no input, we in- DEA and AHP methods is to provide proposing a new aggregation
clude a single input in the last column of Table 6. Now consider technique for the railway station site selection application. This al-
the following DEA model corresponding to the first alternative. lows an alternative formulation for comparing the potential rail-
way stations and selecting the one most preferred. Another
z1 ¼ max 0:213u1 þ 0:067u2 þ 0:150u3 þ 0:108u4 reason is that, a combining DEA and AHP is free rank reversal that
s:t: 0:213u1 þ 0:067u2 þ 0:150u3 þ 0:108u4 6 1 is it does not suffer from rank reversal when an irrelevant alterna-
0:217u1 þ 0:278u2 þ 0:187u3 þ 0:245u4 6 1 tive(s) is added or removed (Ramanathan, 2006).
Although determining the best site for a railway station using
0:218u1 þ 0:331u2 þ 0:268u3 þ 0:278u4 6 1 ð2Þ
AHP and DEA methods has not been discussed so far in the litera-
0:121u1 þ 0:044u2 þ 0:134u3 þ 0:152u4 6 1 ture, we can compare the result of our AHP and DEA methods with
0:230u1 þ 0:279u2 þ 0:260u3 þ 0:217u4 6 1 the method suggested by Ramanathan (2006). For this, we choose
uj P e j ¼ 1; . . . ; 4 the work of Ramanathan (2006) for comparison because it uses the
DEA method for extracting the AHP local and global weights for all
comparisons. Also, the other proposed DEA methods in the litera-
Table 6 ture (Wang & Chin, 2009; Wang et al. 2008) use a similar scheme
DEA observations.
for obtaining the weights or priorities of comparison matrix in the
DMUs O1 O2 O3 O4 Dummy input AHP method. Now let us apply the DEA method of Ramanathan
DMU1 0.213 0.067 0.150 0.108 1 (2006) to the railway station site selection in Mashhad city. First,
DMU2 0.217 0.278 0.187 0.245 1 note that the derivation of local AHP weights for the Mashhad rail-
DMU3 0.218 0.331 0.268 0.278 1 way station site selection using the DEA method proposed in
DMU4 0.121 0.044 0.134 0.152 1
Ramanathan (2006) needs the optimal solutions of 31 DEA models.
DMU5 0.230 0.279 0.260 0.217 1
This is because there are 31 pairwise comparisons in the hierarchy
 N. Mohajeri, G.R. Amin / Computers & Industrial Engineering 59 (2010) 107–114 113

model in the present paper (Fig. 1), i.e., one DEA model for compar- information on the importance of the rail-related subcriteria. The
ing the main criteria with respect to goal, four DEA models for proposed DEA method (Ramanathan, 2006) has the same problem
comparing subcriteria with respect to main criteria, and 26 DEA in the determination of the local weights of remaining comparisons.
models for comparison of alternatives with respect to subcriteria. Consequently, using the proposed DEA method (Ramanathan, 2006)
For each model, Ramanathan (2006) defines the columns of the to determine the local priorities of the potential railway-station
corresponding comparison matrix as multiple outputs and a single sites needs further information as well as solving a series of DEA
dummy input. For example, Table 8, where each rail-related sub- models. Thus, the present paper used the optimal solution of a sin-
criterion is a DMU, shows the element of the comparison matrix gle DEA model for the aggregation of the local AHP priorities among
corresponding to the rail-related subcriteria respect to rail criteria. the exiting railway station alternatives.
For computing the local weight of each subcriterion, a DEA
model must be solved (Ramanathan, 2006). For instance, the 7. Conclusions
importance weight of the first subcriterion, C11, can be obtained
from the following DEA model. In this paper we investigate the problem of finding the opti-
max u1 þ 2u2 þ 3u3 þ u4 þ 3u5 þ 2u6 þ 3u7 mum location for Mashhad railway station site. We propose a hier-
archy model consisting of four main criteria including rail-related,
s:t: u1 þ 2u2 þ 3u3 þ u4 þ 3u5 þ 2u6 þ 3u7 6 1
passenger services, architecture and urbanism, and economics cri-
1 teria. We define local AHP priority weights as multiple outputs and
u1 þ u2 þ 3u3 þ u4 þ 3u5 þ 2u6 þ 4u7 6 1
2 suggest a DEA model for aggregation of the optimum location of
1 1 railway station. The contribution of this paper can be summarized
u1 þ u2 þ u3 þ 3u4 þ u5 þ u6 þ 2u7 6 1
3 3 as follows:
1 This paper proposes a solution to the problem of finding the
u1 þ u2 þ u3 þ u4 þ 3u5 þ 2u6 þ 3u7 6 1
3 optimum location for the Mashhad railway station site. However,
1 1 1 the results are quite general and should be applicable when choos-
u1 þ u2 þ u3 þ u4 þ u5 þ 2u6 þ u7 6 1
3 3 3 ing the optimum location of various types of facilities, particularly
1 1 1 1 in cities, subject to multiple evaluation criteria. Using a new inte-
u1 þ u2 þ u3 þ u4 þ u5 þ u6 þ 2u7 6 1
2 2 2 2 gration of the AHP and DEA methods, in the paper shows how
1 1 1 1 1 the most suitable site for a railway station can be found. The re-
u1 þ u2 þ u3 þ u4 þ u5 þ u6 þ u7 6 1
3 4 2 3 2 sults are promising and show the advantages of the proposed com-
ur P 0 r ¼ 1; . . . ; 7 bined DEA and AHP method. This combination does not suffer from
rank reversal when an irrelevant alternative(s) is added or re-
where, ur ; r ¼ 1; . . . ; 7 denotes the decision variables of the above
moved and provides an alternative formulation for comparing
DEA model. The optimal value of the above model is z1 ¼ 1. Also,
the potential railway stations and selecting the most preferred
if we solve the DEA models corresponding to the other subcriteria
station.
we obtain the following scores as the local AHP weights.
z1 ¼ z2 ¼ z3 ¼ z4 ¼ z5 ¼ 1; z6 ¼ 0:6; z7 ¼ 0:333 Acknowledgments

Now we compare the above local weights, obtained by the DEA

The authors thank to two anonymous reviewers and the editor
method (Ramanathan, 2006), with those obtained in this paper
of CAIE for many helpful comments on the original version and
shown in Table 3a. According to Table 3a the local AHP priorities
Professor Agust Gudmundsson for improving the English of the
corresponding to the rail-related subcriteria are as follows:
second and third revised versions.
wðC 11 Þ ¼ 0:15; wðC 12 Þ ¼ 0:166; wðC 13 Þ ¼ 0:089; wðC 14 Þ
References
¼ 0:239; wðC 15 Þ ¼ 0:101; wðC 16 Þ ¼ 0:175; wðC 17 Þ ¼ 0:08
American Planning Association. (2006). Planning and urban design standards. USA:
where, wðC 1i Þ; i ¼ 1; . . . ; 7 is the local priority of the ith rail-related John Wiley and Sons, Inc.
subcriterion (i = 1, . . . , 7) obtained in the present paper. The above American Railway Engineering and Maintenance-of-way Association (AREMA).
(2006). Manual for railway engineering: Track/structures/infrastructure and
values indicate that all of the rail-related subcriteria have different passenger/systems management/portfolio of track way plans. USA.
priorities. For example, C14 has the highest priority among the rail- Amin, G.R., & Emrouznejad, A. (2010). Finding relevant search engines results: A
related subcriteria. This is followed by C16, C12, C11, C15, C13, and C17. minimax linear programming approach. Journal of the Operational Research
Society, doi:10.1057/jors.2009.53.
But the local priorities obtained from the DEA method (Ramana- Ballis, A. (2003). Airport site selection based on multicriteria analysis: The case
than, 2006) for rail-related subcriteria give the same priority for five study of the island of Samothraki. Operational Research, an International Journal,
subcriteria, C11, C12, C13, C14, and C15. This difficulty may be removed 3(3), 261–279.
Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some models for estimating
by imposing additional restrictions into the corresponding DEA technical and scale inefficiencies. Management Science, 30(9), 1078–1092.
model (Ramanathan, 2006). For doing so, however, needs further Boroushaki, S., & Malczewski, J. (2008). Implementing an extension of the analytical
hierarchy process using ordered weighted averaging operators with fuzzy
quantifiers in ArcGIS. Computers & Geosciences, 34(4), 399–410.
Table 8 Chang, N. B., Parvathinathan, G., & Breeden, J. B. (2008). Combining GIS with fuzzy
Comparison matrix for rail-related subcriteria. multicriteria decision-making for landfill siting in a fast-growing urban region.
Journal of Environmental Management, 87(1), 139–153.
C11 C12 C13 C14 C15 C16 C17 Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of
decision-making units. European Journal of Operational Research, 2(6), 429–444.
C11 1 2 3 1 3 2 3 Cooper, W. W., Seiford, L. M., & Tone, K. (2007). Data envelopment analysis: A
C12 1
2
1 3 1 3 2 4 comprehensive text with models, applications, references and dea-solver software
C13 1 1 1 3 1 1 2 (2nd ed.). USA: Springer.
3 3
C14 1 1 1 1 3 2 3 Dagdeviren, M., Yavuz, S., & Kilinc, N. (2009). Weapon selection using the AHP and
3
TOPSIS methods under fuzzy environment. Expert Systems with Applications,
C15 1 1 1 1 1 2 1
3 3 3 36(4), 8143–8151.
C16 1
2
1
2
1 1
2
1
2
1 2 Emrouznejad, A., Parker, B., & Tavares, G. (2008). Evaluation of research in efficiency
C17 1 1 1 1 1 1 1 and productivity: A survey and analysis of the first 30 years of scholarly
3 4 2 3 2
literature in DEA. Socio-Economic Planning Sciences, 42(3), 151–157.
114 N. Mohajeri, G.R. Amin / Computers & Industrial Engineering 59 (2010) 107–114
Forman, E. H., & Gass, S. I. (2001). The analytical hierarchy process—an exposition.
Operations Research, 49(4), 469–487.
Garcia-Cascales, M. S., & Lamata, M. T. (2009). Selection of a cleaning system for
engine maintenance based on the analytic hierarchy process. Computers &
Industrial Engineering, 56(4), 1442–1451.
Guiqin, W., Li, Q., Guoxue, L., & Lijun, C. (2009). Landfill site selection using spatial
information technologies and AHP: A case study in Beijing, China. Journal of
Environmental Management, 90(8), 2414–2421.
Hsu, T. K., Tsai, Y. F., & Wu, H. H. (2009). The preference analysis for tourist choice of
destination: A case study of Taiwan. Tourism Management, 30(2), 288–297.
Jyrki, W., Dyer, J. S., Fishburn, P. C., Steuer, R. E., Zionts, S., & Deb, K. (2008). Multiple
criteria decision making, multiattribute utility theory: Recent accomplishments
and what lies ahead. Management Science, 54(7), 1339–1340.
Korpela, J., Lehmusvaara, A., & Nisonen, J. (2007). Warehouse operator selection by
combining AHP and DEA methodologies. International Journal of Production
Economics, 108(1–2), 135–142.
Lindebo, E. (2004). Trends in the economic capacity of the Danish fishing fleet,
1996–2002. Acta Agriculturae Scandinavica, Section C – Economy, 1(4),
207–221.
Linkov, I., Satterstrom, F. K., Steevens, J., Ferguson, E., & Pleus, R. C. (2007). Multi-
criteria decision analysis and environmental risk assessment for nanomaterials.
Journal of Nanoparticle Research, 9(4), 543–554.
Managi, S., & Karemera, D. (2004). Input and output biased technological change in
US agriculture. Applied Economics Letters, 11(5), 283–286.
Nijkamp, P. (2004). Transport systems and policy. USA: Edward Elgar Publishing, Inc.
Onut, S., & Soner, S. (2008). Transshipment site selection using the AHP and
TOPSIS approaches under fuzzy environment. Waste Management, 28(9),
1552–1559.
Raharjo, H., Xie, M., & Brombacher, A. C. (2009). On modeling dynamic priorities in
the analytic hierarchy process using compositional data analysis. European
Journal of Operational Research, 194(3), 834–846.
Ramanathan, R. (2006). Data envelopment analysis for weight derivation and
aggregation in the analytic hierarchy process. Computers & Operations Research,
33(5), 1289–1307.
Rosenberg, J. L., & Esnard, A. M. (2008). Applying a hybrid scoring methodology to
transit site selection. Journal of Urban Planning and Development, 134(4),
180–186.
Ross, J. (2001). Railway station, planning, design and management. USA:
Architectural Press (illustrated edition).
Saaty, T. L. (1980). The analytic hierarchy process: Planning, priority setting, resource
allocation. New York, USA: McGraw-Hill.
Saaty, T. L. (2008). Relative measurement and its generalization in decision making:
Why pairwise comparisons are central in mathematics for the measurement of
intangible factors, the analytic hierarchy/network process. Review of the Royal
Spanish Academy of Sciences, Series A, Mathematics, 102(2), 251–318.
Vahidnia, M. H., Alesheikh, A. A., & Alimohammadi, A. (2009). Hospital site selection
using fuzzy AHP and its derivatives. Journal of Environmental Management,
90(10), 3048–3056.
Wang, Y. M., & Chin, K. S. (2007). Discriminating DEA efficient candidates by
considering their least relative total scores. Journal of Computational and Applied
Mathematics, 206(1), 209–215.
Wang, Y. M., & Chin, K. S. (2009). A new data envelopment analysis method for
priority determination and group decision making in the analytic hierarchy
process. European Journal of Operational Research, 195(1), 239–250.
Wang, Y. M., Liu, J., & Elhag, T. M. S. (2008). An integrated AHP–DEA methodology
for bridge risk assessment. Computers & Industrial Engineering, 54(3), 513–525.