Then a spanning tree of graph G can be ex-
pressed by the vector x. Let X be the set of all such
vectors corresponding to spanning trees in graph
G, the mc-MST problem can be formulated as:
min z
1
x
X
m
i1
w
1i
x
i
min z
2
x
X
m
i1
w
2i
x
i
. . .
min z
p
x
X
m
i1
w
pi
x
i
s:t: x 2 X ;
1
where z
i
x is the ith objective to be minimized for
the problem.
Compared with the traditional MST problem,
the problem (1) only diers in the number of ob-
jectives. However, it is because these multiple ob-
jectives exist and usually con¯ict with each other,
we cannot determine which edge has the least
weights and span one by one to form a spanning
tree like the process of vertex growth or edge
growth. If we transform the multiple objectives
into a single objective according to some criteria,
we really can easily solve the problem by using any
MST algorithm but only one single solution in the
sense of Pareto optimality can be obtained. Actu-
ally, this transformation is not an easy work both
for the decision maker and the system analyzer in
practice.
On the other hand, the problem (1) can be
transformed into a multiple objective integer pro-
gramming and then we can make use of the tech-
niques in multiple objective integer programming
to cope with [11]. But with the increase of the
problem scale (vertices), it is still dicult to deal
with. Moreover, the transformation neglects the
topological characteristic of a tree structure in
network and makes the problem even more com-
plicated. Therefore, it is necessary to develop a
new approach to cope with this multi-criteria
problem while not neglecting its tree structure and
give out enough alternatives for the decision
maker to choose. In this paper the GA is adopted
as the new approach for this problem because the
GA can handle any kind of objective functions and
any kind of constraints as an EC method. Before
discussing the GA approach on the mc-MST
problem, some techniques in multiple criteria de-
cision making have to be discussed.
3. Multiple criteria decision making
A basic understanding of MCDM can be il-
lustrated by the following de®nitions [5].
De®nition 1. Given a set of feasible solution
S fx j x 2 X g, solution x
is denoted as the
dominating solution for the problem (1) if and only
if all other solution x 2 S, the following conditions
hold:
z
k
x P z
k
x
; k 1; 2; . . . ; p:
In most cases of MCDM, such a dominating
solution does not exist because these multiple ob-
jectives usually con¯ict with each other and no one
can dominate others. However, if we only consider
each objective respectively, we can ®gure out an
ideal optimal solution.
De®nition 2. As to the problem (1), the point
z
0
x z
0
1
x; z
0
2
x; . . . ; z
0
p
x in criteria space is
denoted as the ideal point, where
z
0
k
x min
x2X
z
k
x; k 1; 2; . . . ; p:
In the MST with single objective, the ideal
point is the dominating solution or optimal solu-
tion. But in the mc-MST problem, actually the
ideal point does not exist because we cannot ob-
tain the solution corresponding to the ideal point.
For the multi-criteria problem, people usually
adopt the concept of nondominated solution or
Pareto optimal solution to de®ne its solution.
De®nition 3. Given a set of feasible solutions
S fx j x 2 X g, solution x
0
is denoted as the
nondominated solution or Pareto optimal solution
for the problem (1) if and only if there is no other
solution x 2 S, satisfying the following condi-
tions:
G. Zhou, M. Gen / European Journal of Operational Research 114 (1999) 141±152 143