V. Mousseau, R. Slowinski & P. Zielniewicz
ELECTRE TRI is a Multiple Criteria Sorting Method ([Yu 92], [Roy, Bouyssou
93]) which aims at assigning a set of alternatives A={a1, a2, ..., al}
evaluated by n criteria g1, g2, ..., gn to one of the preference-ordered
and pre-defined categories. The assignment of an alternative ak to a specific
category results from a comparison of its evaluation on all criteria to
the profiles defining the limits between consecutive categories.
We present in this section a new implementation of ELECTRE TRI. It
integrates specific functionalities supporting the Decision Maker (DM)
in the preference elicitation process. These functionalities grouped in
a module called ELECTRE TRI Assistant aim at reducing the cognitive effort
required from the DM in the phase of calibration of the preference model.
The main characteristic feature of ELECTRE TRI Assistant is the inference
of the ELECTRE TRI preferential parameters from assignment examples supplied
by the DM.
The ELECTRE TRI software version 2.0a (Mousseau et al. 1999, 2000)
is the result of a collaboration between LAMSADE (University of Paris-Dauphine,
France) and the Institute of Computing Science (Poznan University of Technology,
Poland). It is developed in C++ programming language and runs under Microsoft
Windows 3.1, 95, 98, NT.
The structure of the options available in the software is the following:
Support for Parameters Elicitation: ELECTRE TRI Assistant
One of the main difficulties that an analyst must face when interacting
with a DM in order to build a decision aid procedure is the elicitation
of various parameters of the DM's preference model. In the ELECTRE TRI
method, the analyst should give values to profiles, weights and thresholds.
Even if these parameters can be interpreted, it is difficult to fix directly
their values and to have a clear global understanding of the implications
of these values on the output of the model.
[Mousseau, Slowinski 98] proposed a methodology that avoids this problem
by substituting assignment examples for direct elicitation of the model
parameters. The values of the parameters are inferred through a certain
form of regression on assignment examples. ELECTRE TRI Assistant implements
this methodology in a way that requires from the DM much less cognitive
effort: the elicitation of parameters is done indirectly using holistic
information given by the DM through assignment examples, i.e. alternatives
assigned by the DM to categories according to his/her comprehensive preferences.
The ELECTRE TRI Software 2.0 includes an assistant that is able to
infer preference model parameters from assignment examples provided by
the user. The present version supports the user in defining the weights
of criteria and the cutting level l for the pessimistic assignment procedure
only (the next version will include similar functionalities for profiles
and thresholds). The use of ELECTRE TRI Assistant functionalities proceeds
according to the following scheme:
(a) input the list of assignment examples composed of alternatives
for which the DM gives a holistic assignment (such alternative can be an
existing alternative of a fictitious one designed for this purpose); imprecise
assignments are accepted, i.e, the DM can express an hesitation in the
assignment of an alternative a by specifying a subset of consecutive categories
to which a could be assigned,
(b) give preferential information on the weights and/or cutting level
(preorder, comparisons of specific coalitions, bounds on weights, on the
cutting level l) - optional,
(c) run the inference procedure to find the most adequate values of
the weights,
(d) check for the acceptability of the obtained weight vector and,
either:
Contacts:
Lamsade Softwares, Université Paris Dauphine,
Place du Mal de Lattre de Tassigny
75775 Paris Cedex 16, France
Electre.Tri@lamsade.dauphine.fr
www.lamsade.dauphine.fr/logiciels
Institute of Computing Science,
Poznan University of Technology,
Piotrowo 3A,
60-965 Poznan, Poland
Piotr.Zielniewicz@cs.put.poznan.pl
www.cs-idss.put.poznan.pl
References: