Professional Documents
Culture Documents
Jablonsky SANNA Apli
Jablonsky SANNA Apli
Introduction
Multicriteria evaluation of alternatives belongs among the basic decision problems of
multicriteria decision making with very large possibilities of real applications (evaluation of
investment alternatives, evaluation of credibility of bank clients, rating of companies,
consumer goods evaluation and many others). The theory of multicriteria evaluation of
alternatives is very good established and there are available many different methods for this
kind of problems. From the decision makers point of view the wider utilisation of
multicriteria evaluation techniques depends on the availability of the powerful and userfriendly oriented software tools. The paper describes simple application for multicriteria
evaluation of alternatives developed under MS Excel environment as standard add- in
application written in Visual Basic for Application (VBA). The name of the system is
SANNA (System for ANalysis of Alternatives).
The multicriteria evaluation of alternatives problem is usually defined by criterion matrix
as follows:
Y1
X1
X2
M
Xn
y11
y
21
M
y n1
Y2
K Yk
y12 K y 1k
y 22 K y 2 k
y n 2 K y nk
The paper is supported by the Grant Agency of Czech Republic grant no. 402/98/1488 and corresponds to the
research program of the Faculty of Informatics and Statistics no. CEZ:J18/98:311401001.
94
technique and information that must be specified before or during the evaluation process. The
basic objectives of the evaluation of alternatives are:
selection of the best alternative,
complete ranking of alternatives,
classification of the alternatives (splitting of the alternatives into several classes, e.g. ABC
analysis),
selection of the subset of good alternatives.
The SANNA system does not contain an expert system that could help to decision makers to
select the best solution technique for solving their problems depending on the objectives of
the evaluation. However, the evaluation methods containing in the SANNA system belong
among the standard and most often used techniques and can usually be used in most cases.
All the decision methods supported by SANNA system expect the knowledge of the
information about the decision makers preferences of the criteria in the quantified form. This
means that the decision maker must input the weights of the criteria w1 , w2 , ..., wk . The
specification of the weights is facilitated by the special module of SANNA that helps to
estimate the weights by simple well-known methods.
the new SANNA menu item is added into the basic Excel menu bar. Simultaneously the
SANNA toolbar with buttons for easy working with the application is opened (see Fig. 2).
The structure of the SANNA menu corresponds to the structure of the application. The
SANNA system consists of five basic modules plus simple help tool (each module
corresponds to one item of SANNA menu):
1. Data management module.
SANNA works with standard .xls worksheets. Input data information of the decision
problem with the criterion matrix is placed in this worksheet in the sheet named data.
Data management module either opens new data sheet (if does not already exists) based on
specification of the basic parameters of the problem (the number of alternatives and
criteria) or deals with the current data sheet. The basic functions for dealing with the
95
current data set are adding of the new alternative or criteria or removing the alternative or
criteria marked by the active cell of the sheet.
2. Non-dominance filter.
This module makes it possible to test the non-dominance of alternatives in the current data
set. The alternatives are marked according to the results of the test as dominated or nondominated. Another feature of the module is the possibility to remove the dominated
alternatives from the data set. The removing of dominated alternatives is non-reversible.
3. Estimation of weights module.
The decision maker can specify the weights of the criteria either directly by their setting in
the data sheet or by means of three simple well-known methods supported by estimation of
weights module. The decision maker can use the following techniques:
- scale method (weights are set in any numerical scale and normalised),
- Fuller method (based on paired comparisons of the criteria and selection one of the
following: one of the criteria is more important or both the criteria are equally
important),
- AHP (based on the standard Saatys scale) uses either eigenvector calculation or
logarithmic least square method (approximation of the eigenvector); the consistence of
the comparison matrix is checked.
The estimation weights module contains the function that transmits the weights calculated
by above mentioned methods into the data sheet and the function than graphically presents
the weights.
4. Methods for multicriteria evaluation of alternatives.
The current version of the SANNA system supports five basic methods for multicriteria
evaluation of alternatives (WSA, TOPSIS, ELECTRE I, PROMETHEE I, II and
MAPPAC). All the methods need the knowledge of the weights of the criteria. ELECTRE
and PROMETHEE class methods require specification of some additional parameters. A
brief characteristics of the available methods is following:
WSA
The WSA method is based on linear utility function. This method computes the global
utility of the alternatives as the weighted sum of no rmalised criterion values. The method
provides complete ranking of alternatives according to their global utilities.
96
TOPSIS
The TOPSIS method is based on the computation of global utilities of alternatives
according to their closeness to the ideal criteria values and distance from the nadir criteria
values. Analogous to the WSA the TOPSIS method provides complete ranking of
alternatives according to their global utilities.
ELECTRE I
The basic output of the ELECTRE I method is the split of the set of alternatives into two
disjoint subsets of efficient and inefficient alternatives. Except the criterion matrix and
weights of the criteria the user must specify a concordance and discordance thresholds. The
final result of the evaluation (subset of efficient and inefficient alternatives) is often very
sensitive on the above mentioned thresholds. The user can repeat the calculation with
different sets of thresholds and compare the results.
PROMETHEE I and II
The PROMETHEE methods need the knowledge of criterion matrix, weights of the criteria
and preference functions of criteria with their parameters for measuring the strength of the
preference of the pairs of alternatives with respect to the given criterion. SANNA uses the
same types of preference functions as proposed by authors of the method in [1].
PROMETHEE I method provides partial ranking of alternatives, PROMETHEE II method
offers complete ranking according to the net flow values.
MAPPAC
The MAPPAC method works with the criterion matrix and weights of the criteria only.
The method splits the alternatives into several preference groups. The evaluation of
alternatives by this method can be very long for problems with several tens of alternatives
comparing to other supported approaches.
5. Report wizard.
The successfully completed evaluation by any of the methods adds a new sheet into the
active worksheet with basic information about the results. By report wizard the user of the
application can build his own output sheet with the results in the user-defined form. The
97
user selects required items for the report from the set of items associated to the current
method (including simple graph presentations of results). The basic window of the report
definition for TOPSIS method is presented on Fig. 3.
6. Help.
SANNA contains simple help window that can be activated from the SANNA menu or by
the appropriated button on the SANNA toolbar.
98
sheet that is added in to the worksheet (this sheet is named in accordance with the name of
the method used). After the weights are calculated they can be transmitted into the data
sheet by command Weights-Apply weights. For this command the sheet with results of
weights calculation must be active. Weights contained in the data sheet are used as
inputs for all of the available multicriteria evaluation methods.
99
References
1. Brans,J.P, Mareschal,B.: The PROMCALC and GAIA decision support system for
multicriteria decision aid. University of Brussels, 1991.
2. Fiala,P., Fbry,J.: SYMCA System for Multicriteria Analysis. CEJORE 2(1993), no.2,
pp.177-183.
3. Jablonsk,J.: Multicriteria modelling by means of the LINGO system. In: Methods and
applications of operational research II, Katowice 1998, pp. 63-71.
100