H.J.M. Fernhout
Methods for decision analysis: a basic introduction.
Literature survey,
Report 2003.TL.6750, Transport Technology, Logistic Engineering.
Multi Criteria Decision Analysis (MCDA) is often used to aid the decision
making process. Since there is diversity in the problems that need
analysis, there is also diversity in the methods that are used for this
analysis.
The Multi Attribute Value Theory (MAVT) is the most widely used method for
MCDA. In most of the situations use of SMART as a weighting method for
MAVT is preferred, at least initially; the use of the 'easiest' weighting
method is always preferred, if it does not effect the outcome of the
decision. A strong side of MAVT is the simplicity of the application. The
(main) weakness of MAVT is the lack of analytic support for ensuring
consistency of judgments.
The AHP-method (Analytical Hierarchy Process) is
also widely used for MCDA. The weakness of AHP is the non-intuitive
interpretation of weights and the dramatic growth in the number of
comparisons, if there are a lot of alternatives and criteria. A major
advantage of AHP is that the decision maker is forced to structure the
problem hierarchically; because of this the decision maker usually looks
at a problem from a different perspective, making the process worthwhile
without regards to outcome of the process. Eigenvalue-calculation is an
effective method for detecting inconsistencies in the preference relation of
the alternatives and criteria. There are more AHP-like methods.
The main purpose of variants to existing methods is the fact that the standard
methods do not always get the job done. Sometimes a better result will be
achieved, or a much easier analysis, just by using one of the many variants.
Fuzzy (logic) is a method that is somewhat similar to AHP; it uses fuzzy sets
to describe uncertain input values. A fuzzy set is kind of a distribution.
The Delphi method, although mainly used in forecasting events or
breakthroughs (e.g. scientific), can be used for variety of different
applications. Some of those are: planning and budget evaluation. Some
key-features of the Delphi Method are: panel use and anonymity of the
participants. An advantage of the Delphi method over a 'normal' meeting is
the fact that the participants do not have to be at the same location and
do not have to participate simultaneously. There is also critique on the
Delphi method, some common reasons for failure are: assuming Delphi to be a
surrogate for human communication and underestimating the demanding nature of
a Delphi.
There are a lot of variations to and variants of existing methods. The
variants discussed are used at the TU Delft. The described variants are:
(1) Use of Knock-Out-ratings, which would disqualify the concerned
alternative. (2) Use of weighting-scenarios, which represent different
opinions/preferences. (3) An AHP-like method, with MAVT characteristics,
that compares to an ideal-score.
The real challenge might even be finding the best-suited method. But on
the other side: any method that produces the correct results is a correct
one. To distinguish 'good' from 'bad' results the expertise of a (or
multiple) good decision maker(s) is required; the method does not stand
above him/her, it is just a tool.
Reports on Logistic Engineering (in Dutch)
Modified: 2003.04.09;
logistics@3mE.tudelft.nl
, TU Delft
/ 3mE
/ TT
/ LT.