A GOAL-ORIENTED EVALUATION OF

LOCATION/ALLOCATION METHODS

Nazila Mohammadi

Mohammad Reza Malek

Ali Asghar Alesheikh

Dept. of GIS, Faculty of Geodesy and Geomatics Eng., K.N. Toosi University of
Technology, Tehran, Iran.

[email protected] [email protected]
[email protected]

Abstract. The main purpose of any facility location is to select the optimal places that
satisfy project goals. In location problems, the object is usually to optimize a function
that is called objective function. The function defines the problem conditions and
efficient decision parameters. Numerous methods are proposed to challenge the location
facility issues. Each of them is appropriate for solving one type of problems according to
its constraints.
In the framework of this research, facility location methods and problems are classified
into four distinct categories based on their goals. Each class is then formulated as an
optimization problem. In order to evaluate the efficiency of each method, a case study is
considered. The methods are evaluated based on the data obtained from study area. The
implementation results are compared quantitatively. In conclusion, application of each
method was examined and presented based on the problem’s goals.

Keywords: Facility location/allocation, Optimization, Classification, GIS,

objective function

1 Introduction

Facility location selects the optimal location with maximum efficiency according to the
goals of projects. Solving location problems needs to consider the effective constraints
and criteria in its computational process [1]. Recent decades witness huge improvements
in computational research and information technology, and as such the application of
Geospatial Information System (GIS) in facility location is increased exponentially [2].
Spatial scientists are continuously working on mathematical modeling and solution of
problems concerning the placement of facilities to minimize transportation costs, prevent
placing hazardous materials near housing, outrun competitors' facilities, etc. Several
objective functions relevant to each model, are also formulated [3]. In location problems,
all input values such as demand rate, distances and time of service transition are
considered as instant parameters and then outputs are estimated as decision variables in
one point of time [4].
The systematic study on location problem formally began by Weber in 1909. He
considered how to position a single warehouse to minimize total distance between it and
several customers 5]. Since the mid 1960s, the study of location theory has been
developed and basic problems in this context have been formulated as static and
deterministic problems [6].
Static problems deal with constant known quantities as input. According to the
problem constraints, the processing results give out a single solution to implement at one
point in time. Following this, in 1964 Hakimi extended location theory with his
researches to locate switching centers in a communications network in a highway [7].
Some of the problems require more facility centers to be located. Thus, multisource
Weber problem has been defined [8]. The multisource Weber problem is defined as
locating simultaneously m facilities in the Euclidean plane in order to minimize the total
transportation cost for satisfying the demand of n users, each supplied from its closest
facility [9].
Location problems can be solved through various strategies. Determining suitable
solutions is difficult for users. Without a general solution classification, the examination
and comparison of the methods maybe difficult if not impossible.
Accordingly, within the framework of this paper a general classification of location
methods and their mathematical formulations are presented. The classifications are done
based on the categories fitness functions and their applications in solving the problems .
In order to compare the efficiency and the usage of each method, simulated data are
generated. Models are classified into four categories. Software has been developed to test
each category. The results are then compared quantitatively. At last, application scenarios
that best suit each category are identified.
The rest of this paper is organized as follows: Section 2 describes how to classify the
different location methods. Section 3 explains how to compare the methods by
optimization problems. Finally, we give concluding remarks in section 4.

2 Classification and Mathematical Formulation

Facility location models are used in a wide variety of applications. These include, but are
not limited to, locating warehouses, hazardous materials, public and private centers,
rescue stations. Each of these location problems can generally solved on one of three
basic continuous, discrete and network spaces. They all have different objective functions
according to problem details. From a classification point of view, facility location models
are different in their objective functions.
According to wide extent and diversity of problems that related to location and special
characteristics of each location problem, several location methods have been presented
for solving each of them. Apart from details of private constraints and conditions refers to
each of different problems, almost all location methods can be formulated as a few
general distinct classes according to their application and their applied fitness function.
Advantage of this general classification is in possibility of presenting a unique
formulation for problems with similar constrains. Thus with existing differs in details,
general prospect in method of solving problem is obtained. Therefore, applying
constraints and conditions to the general formulae of the selected appropriate general
method leads to the desired specific problem. This section provides classification and
mathematical formulation of various location methods.

2.1 Median problem

First class is known as p-median class. P-median problem is a version of the multisource
Weber problem [7]. The purpose of this problem is to find the location of p facilities to
minimize total demand-weighted travel distance between demands and facilities [6].
For mathematical formulation of these problem decision variables defines as follow:
⎧1 If it locates at potential facility site j
Xj =⎨
⎩0 Otherwise

⎧1 If demands at node i are served by a facility at node j

Yij = ⎨
⎩0 Otherwise

In the above equations i is index of demand point and j is index of potential facility
site. Accordingly, p-median problem can be formulated as (1) to (6).

Minimize∑ ∑ hi d ij Yij
(1)
∑X j =1
j (2)
∑Y ij =1
j ∀i (3)
Yij − X j ≤ 0 ∀i, j
(4)
X j ∈ {0,1} ∀j
(5)
Yij ∈ {0,1} ∀i, j
(6)

Where hi is demand at nod i, dij expresses the distance between demand node i and
potential facility site j and p is number of facilities to be located.
As it can be seen in the aforementioned formula, the objective function (1) minimizes
the total demand-weighted distance between facility centers and demand nods. The
constraints (2) control that exactly p facilities be located. Equations (3) ensure that all
demands refer to centers. The family constraint (4) allows assignment only to sites which
facilities have been located. The family constraint (5) and (6) are binary requirements for
the problem variables. With considering six above formulas together, solving this
problem is leads to solve an optimization problem that needs to minimize (1) so that
satisfy constraint (2) to (6).
2.2 Covering problems

A demand is said to be covered if it can be served within a specified time or distance [8].
Covering problems can be divided into two main parts. One is related to that in which
coverage is required and another is related to that in which is optimized. The former is
called covering problem and later is called maximal covering problem [13].

2.2.1 Set covering problems

In the set covering problem, the objective is to select appropriate locations for facility
centers so that minimize the cost of satisfying demands such that a specified level of
coverage is obtained. In this problem coverage distance is fixed and the minimum
number of facility centers hat is necessary for covering the all demands, is determined.
By the set covering problem one can examine how many facilities are needed to
guarantee a certain level of coverage to all users.
This problem can be formulated by equation systems (7) to (9).

Minimize∑ c j X j (7)
j

∑X j ≥1 (8)
j∈N i ∀i
X j ∈ {0,1} ∀j (9)

In above equations, cj is fixed cost of sitting facility at node j, S is maximum

acceptable service distance or time and Ni is the set of facility sites j within acceptable
distance of node i.
The cost of facility location is minimized by the objective function (7). The family
constraint (8) explains that all demand i have at least one facility located within the
acceptable service distance. Constraint (9) insures integrity for the decision variables.
This formulation makes no different between demand nodes. All of the nodes are
assumed that have an equal demand rate and must be covered within the specified
distance.

2.2.2 Maximum covering problem

Maximum covering problem seeks to maximize the amount of demand covered within
the acceptable service distance by locating a fixed number of facilities. For mathematical
formulation of this problem, decision variable Zi defines as follows:
⎧1 if node i is covered
Zi = ⎨
⎩0 Otherwise

This problem can be formulated by the formulae (10) to (14).

Maximize∑ hi Z j (10)
i
Zi ≤ ∑X j (11)
j∈N i ∀i
∑X
j
j ≤p (12)
X j ∈ {0,1} ∀j (13)
Z i ∈ {0,1} ∀i (14)

The objective function (10) maximizes the amount of demand points covered.
Inequalities (11) define which demand nodes are covered within the fixed distance
covered. Constraint (12) controls the number of facility centers and limits it to fixed
number. Similar to (9), the constraints (13) and (14) are integrity constraints for the
decision variables.
The relationship between the p-median problem and maximum covering problem can
be found by defining d'ij [14-15] as follows:
⎧0 if d ij ≤ S
d ij′ = ⎨
⎩1 if d ij > S
Using d'ij in p-median equations, the amount of no served demands within coverage
distance S will be minimized. This means to maximize the amount of demands served
within S. With this transformation, maximum cover problem can be viewed as a spatial
case of the p-median problem with an additional parameter S.

2.3 Center problems

The p-center problem in a mathematical framework is known as a MinMax problem. That

means the aim is to minimize the maximum distance between any demands and is nearest
facility. If D is maximum distance between a demand node and he nearest facility, p-
center problem can be formulated as equation (15) with constraints (16) to (21).
The object function (15) minimizes the maximum distance between any demand node
and its nearest facility. Constraints (16) to (18) are similar to constraints (2) to
(4).equation (19) determines the maximum distance between any demand node and the
nearest facility. Constraints (20) and (21) are integrality constraints for decision
variables. In case of decision variables Yij could be fractional; one demand node might
be served by multiple facilities.

Minimize D (15)
∑X
j
j =p (16)

∑Y ij =1 (17)
j ∀i
Yij − X j ≤ 0 ∀i, j (18)
D ≥ ∑ d ij Yij (19)
j ∀i
X j ∈ {0,1} ∀j (20)
Yij ∈ {0,1} ∀i, j (21)

3 Simulation

In order to compare the efficiency of classified methods in solving a unique problem, a

section of central part of Tehran, the capital city of Iran, with nineteen hospitals has been
chosen (figure 1).
The primary purpose is to find the best place for establishing a catering center to
satisfy the existing hospitals demands. Demands rate has been assumed equal for all
hospitals. Euclidean distance has been considered as the only efficient parameter in
satisfying the demands. Covering distance, in cover problems, of 160 units has been
assumed..
According to the previous equations the best places for establishing catering centers are
determined and presented. (Figure 2). In figure 2, circles indicate facility centers and blue
squares indicate hospitals as demand points that have been referred to related facility
centers with lines.

Figure 1: A section of central part of Tehran, capital of Iran, as sample region

 Figure 2: The best places for establishing facility centers from top and right
respectively are median, center, max-cover and set-cover

Table 1 presents the results of numeric computations.

Table 1: Numeric result of comparing different location problems

Total
demand Maximum
distance distance
Coordinates of Number of
between between any
selected facility covered
facility demand node
centers demand nodes
centers and and its nearest
demand facility
nodes
Median X1=230.413 ,
2745.14843 236.98601 19
problem Y1=153.518
X1=246.564 ,
Center problem 2871.12880 196.56422 19
Y1=102.005
Max cover X1=298.448 ,
1321.59981 141.80786 12
problem Y1=151.269

X1=103.057,
Y1=151.799 10
Set cover
1734.28714 151.23283
problem
X2=351.437 , 9
Y2=144.416

Table 1 presents the location of best facility centers using various methods. As it can
be seen, the locations are not the same. Thus, each problem must be categorized properly,
and then, solved using unique solutions presented in this paper. The solutions must be
tuned according to the application at hand, its specific constraints and goals.
Figure 3 illustrates the position of facility centers on the city map.

Figure 3: Selected locations using each of four methods on sample region

3 Conclusions

This paper addressed the location problems. The problems were classified into four
distinct categories. Each category is examined using simulated data, and the results are
compared quantitatively.
We have demonstrated that in problems such as establishing public facility centers like
school and hospital, p-center method is appropriate. Because in this type of problems the
purpose is to satisfy all demand nodes. Even the discrete nodes that are located far away
must be served similar to other demand nodes. In other words, facility centers must be
located in a place that is close to demand nodes. Thus, p-center problem is more suitable
in this type of problems.
If the problem is to establish emergency centers such as fire stations, we need to define
a coverage distance or time for the services. Thus, the objective is to cover the region
with minimum centers considering a fixed cover distance or time. In this type of
problems, set cover is the best solution.
In problems such as determining the optimum location for establishing markets, p-
median is an appropriate method. In such problems, although discrete demand points are
far from centers but nearing to focused section from the viewpoint of demand nodes, is
economical.
If the problem is related to establishing private facility centers such as restaurants,
maximum cover can be the best choice for solving these problems. Because in
establishing private facility centers, the goal is to serve the maximum demand nodes in a
determinate cover distance.
We demonstrated that through the proposed classification, almost all location problems
can be categorized.
As future researches, authors want to classify and determine mathematical methods for
solving dynamic location problems.

5 References

1. Schilling, D.A.: Dynamic location modeling for public-sector facilities: A multicriteria

approach. Decision Sciences 11, pp.174-724 (1980)
2. Alesheikh, A.A., Soltani, M.J., Nouri, N., Khalizadeh, M.: Land Assessment for Flood
Spreading Site Selection Using Geospatial Information System. International Journal of
Environmental Science and Technology, 5(4), pp. 455-462 (2008)
3. Bischoff, M., Klamroth, K.: Two branch and bound methods for a generalized class of
location-allocation problems. Technical Report No. 313, In stitute of Applied
Mathematics, University of Erlangen-Numberg.
4. Current, J., Min, H., Schilling, D.: Multiobjective analysis of facility location
decisions. European Journal of Operational Research, vol. 49, pp. 295-307 (1990)
5. Weber, A.: Alfred Weber’s Theory of the Locaion of Industries. University of Chicago
(1929)
6. Owen, S.H., Daskin, and M.S.: Strategic facility location: A review. European Journal
of Operational Research, vol. 111, pp. 423-447 (1998)
7. Hakimi, S.L.: Optimum location of switching centers and he absolute centers and
medians of a graph. Operations Research, vol. 12, pp. 450-459 (1964)
8. Bongartz, I., Calamai, P.H., Conn, A.R.: A projection method for norm location-
allocation problems. mathematical programming, vol. 66, pp. 283-312 (1994)
9. Trevors, hale.: Location science research: a review. Annals of operations research, vol.
123, pp. 21-35 (2003)
10. White, J.A., Case, K.E.: On covering problems and he central facilities location
problem. Geographical Analysis, vol. 6, pp. 281-293 (1974)
11. Schilling, D.A., Jayaraman, V. R.: A review of covering problem in facility location.
Location Science, vol. 1, pp. 25-55 (1993)
12. Church, R.L., ReVelle, C.S.: heretical and computational links between the p-median
location set-covering and the maximal covering location problem. Geographical
Analysis, vol. 8, pp. 406-415 (1976)