Multicriteria Flow Shop Scheduling Problem - Enterprise Information Systems Design, Implementation and Management: Organizational Applications - page 221

widely as possible. In the beginning, methods
were adaptations of single-objective optimiza-
tion. Nowadays they have their own entity. They
are initially inspired by EA or neighbourhood
search. Furthermore, recent developments are
more hybridized, given rise to Multi-Objective
Hyperheuristic (MOH) methods. A MOH can be
thought as a heuristic method, which iteratively
selects the most suitable heuristic amongst many
(Burke et al., 2003).
The problem of obtaining a uniformly distrib-
uted set of non-dominated solutions is of great
concern in Pareto optimization. The specification
of the search direction, by tuning weights, is the
method that directly attempts to drive the current
solution towards the desired region of the trade-off
frontier. Hyperheuristic approaches attempt to do
it by applying the neighbourhood search heuristic
that is more likely to drive the solution in the de-
sired direction. This technique can be applied to
single-solution and population-based algorithms.
A priori methods assume that the decision-
maker preferences can be expressed. The hier-
archical approach penalizes too much the less
important criteria, while setting a criterion as
the most important one. In reality, the decision-
maker preferences are usually smooth, giving less
importance to the main criterion and more to the
less important criteria. Considering a composite
function of the criteria involved in the problem,
it is implicitly assumed that the decision-maker
preferences are accurately reflected in this ob-
jective function. The decision-maker knows the
preferable schedule, but it is not easy to express
this preference in a function. In general, a priori
approaches give a solution to the problem that
cannot usually be trusted to be the most preferred
solution.
To be confident with a particular solution to a
problem with multiple objectives, the decision-
maker active involvement is required. In interac-
tive methods, she indicates their preferences during
the process of solution, guiding the search direc-
tion. Agrawal et al. (2008) proposes an interactive

particle-swarm metaheuristic for MOO. The ap-
proach presented by Jaszkiewicz & Ferhat (1999)
can be placed between the a priori and interactive
procedures. They present a method that includes
some interaction with the decision-maker, but
is based on the assumption that decision-maker
preferences are already relatively well defined at
the beginning of the solution process.
For methods that should offer an approxima-
tion of the complete set of efficient solutions, it
is guaranteed that no potential preferable solution
has been eliminated, but the number of efficient
solutions can be overwhelmingly high to warrant
proper examination by the decision-maker ( a
posteriori approaches).
In this chapter, we focus on the flow-shop
problem. We refer he reader to the following
papers for further information on MOCO:

• Ulungu & Teghem (1994) and Ehrgott &

Gandibleux (2002), for general MOCO
theory.

• Landa-Silva et al. (2004) and Jones et al.

(2002), for overviews on metaheuristics.

• Jaszkiewicz (2004), for a comparison of

metaheuristics for bi-criteria optimization.

• Aickelin (1999) and Jaszkiewicz (2004),

for Multi-Objective Genetic Algorithms
(MOGA). For general Multi-Objective
Evolutionary Algorithms (MOEA), to
Zitzler (1999), Coello & Mariano (2002)
and Geiger (2007).

• Serafini (1992), Ulungu (1993), Hapke

et al. (1998) and Loukil et al. (2005), for
Multi-Objective Simulated Annealing
(MOSA).

• Gandibleux et al. (1997), for Multi-

Objective TS.

evaluation of Multi-objective
pareto Methods

For the MOO algorithms, the analysis of perfor-
mance is more complex than for single-objective