Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

European Journal of Operational Research 138 (2002) 247–259

www.elsevier.com/locate/dsw

Rough sets methodology for sorting problems in


presence of multiple attributes and criteria
a,*
Salvatore Greco , Benedetto Matarazzo a, Roman Slowinski b

a
Faculty of Economics, University of Catania, Corso Italia, 55, 95129 Catania, Italy
b
Institute of Computing Science, Poznan University of Technology, 60-965 Poznan, Poland

Abstract

We consider a sorting (classification) problem in the presence of multiple attributes and criteria, called the MA&C
sorting problem. It consists in assignment of some actions to some pre-defined and preference-ordered decision classes.
The actions are described by a finite set of attributes and criteria. Both attributes and criteria take values from their
domains; however, the domains of attributes are not preference-ordered, while the domains of criteria (scales) are
totally ordered by preference relations. Among the attributes we distinguish between qualitative attributes and
quantitative attributes. In order to construct a comprehensive preference model that could be used to support the
sorting task, we consider preferential information of the decision maker (DM) in the form of assignment examples, i.e.
exemplary assignments of some reference actions to the decision classes. The preference model inferred from these
examples is a set of ‘‘if . . . , then . . . ’’ decision rules. The rules are derived from rough approximations of decision classes
made up of reference actions. They satisfy conditions of completeness and dominance, and manage with possible
ambiguity (inconsistencies) in the set of examples. Our idea of rough approximations involves three relations together:
indiscernibility, similarity and dominance defined on qualitative and quantitative attributes, and on criteria, respec-
tively. The usefulness of this approach is illustrated by an example.  2002 Published by Elsevier Science B.V.

Keywords: Rough sets; Sorting; Classification; Multiple criteria decision analysis; Decision rules

1. Introduction (DM’s) preferential information. Very often, this


information has to be given in terms of preference
The main difficulty with application of many model parameters, like importance weights, sub-
existing multiple-criteria decision aiding (MCDA) stitution rates and various thresholds. It is gener-
methods lies in acquisition of the decision maker’s ally acknowledged, however, that people prefer to
make exemplary decisions than to explain them in
terms of the preference model adopted by the an-
* alyst. For this reason, the idea of inferring pref-
Corresponding author.
E-mail addresses: salgreco@mbox.unict.it (S. Greco),
erence models from exemplary decisions provided
matarazz@mbox.unict.it (B. Matarazzo), by the DM is very attractive. The exemplary de-
slowinsk@sol.put.poznan.pl (R. Slowinski). cisions may, however, be inconsistent because of

0377-2217/02/$ - see front matter  2002 Published by Elsevier Science B.V.


PII: S 0 3 7 7 - 2 2 1 7 ( 0 1 ) 0 0 2 4 4 - 2
248 S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259

limited discriminatory power of criteria and be- Table 1


cause of hesitation of the DM. These inconsisten- Example of firms evaluated for assessment of bankruptcy risk
cies cannot be considered as a simple noise or error Firms Return on Location Bankruptcy
that should be corrected or neglected. They can investment (%) risk
convey important information that should be ta- Firm A 5 City X High risk
Firm B 8 City X Medium risk
ken into account in the construction of the DM’s
Firm C 3 City Y Medium risk
preference model. Firm D 3 City X Medium risk
Rough sets theory (RST) deals with the incon-
sistency problem by separation of certain and
doubtful knowledge extracted from exemplary evaluation does not violate the dominance princi-
decisions. The original rough set approach (see, ple, however, because A and C are located in dif-
e.g., Pawlak, 1991; Slowinski, 1992) is not able, ferent cities where conditions of investment may be
however, to discover inconsistencies related to different. Finally, comparing Firm A to Firm D,
consideration of criteria, i.e. attributes with pref- one discovers that D also has overall evaluation of
erence-ordered domains (scales), like return on bankruptcy risk better than A (medium risk vs.
investment, market share, debt ratio. Regular at- high risk), although A has a higher return on in-
tributes, e.g. symptoms, colours, textural features, vestment. In this case, however, the overall eval-
traditionally considered in the rough set method- uation of D and A cannot be explained by different
ology, are different from criteria because their location because D and A are located in the same
domains are not preference-ordered. The presence city, so this case violates the dominance principle.
of criteria requires consideration of dominance Consequently, Firms A and D are an inconsistent
relation in data analysis. We say that action x pair of examples.
dominates action y if: Let us observe that the above inconsistency will
1. x is at least as good as y on all considered crite- not be detected either by the original rough set
ria and approach or by other methods of data analysis
2. x and y have identical or similar description on that do not take into account the preference order
all considered attributes. in the domains of attributes (criteria) and among
According to rational behaviour, assignment of decision classes. In the context of the original
actions to preference-ordered classes should re- rough set approach, Firm A is clearly discernible
spect the following dominance principle: if action x from Firm D, because they have different de-
dominates action y, then x should be assigned to a scription by attributes, and this legitimates their
class not worse than y. When actions x and y do different assignment to decision classes. So, no
not satisfy the dominance principle, we say that inconsistency will be stated.
the pair ðx; yÞ is inconsistent. To deal with preference-ordered domains of
Consider a simple example of firms evaluated criteria and preference-ordered decision classes,
for assessment of bankruptcy risk. The firms de- Greco et al. (1996, 1998a, 1999) have proposed an
scribed by one criterion (return on investment) and extension of the RST. This innovation is based on
by one regular attribute (location) are assigned by substitution of the indiscernibility relation by a
overall evaluation to ordered classes of bank- dominance relation in the rough approximation of
ruptcy risk. The assignment examples are shown in decision classes. An important consequence of this
Table 1. fact is a possibility of inferring from exemplary
Let us remark that Firm B has a better overall decisions the preference model in terms of decision
evaluation (medium risk) than Firm A (high risk), rules, being logical statements of the type ‘‘if . . . ,
which is rational because B has a higher return on then . . . ’’, with a specific syntax proposed by the
investment than A while their location is the same. authors. The extended RST is also able to deal
Firm C has also a better overall evaluation than with inconsistencies, like the ones presented in
Firm A (medium risk vs. high risk), although A Table 1, which are typical to exemplary decisions
has a higher return on investment. This overall in MCDA problems. The separation of certain and
S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259 249

doubtful knowledge about the DM’s preferences is hesitation of the DM and unstable character of
done by distinction of different kinds of decision her/his preferences.
rules, depending on whether they are induced from 4. For a given set of decision examples, it is able to
lower approximations of decision classes or from determine what criteria and/or attributes are
the boundaries of these classes composed of in- relevant for approximation of the assignment
consistent examples. Such a preference model is made by the DM; such subsets of relevant crite-
more general than the classical functional or rela- ria and/or attributes are called reducts and their
tional model in MCDA and it is more under- intersection is called the core.
standable for the users because of its natural 5. It permits one to handle criteria and attributes
syntax (Greco et al., 2001b). A survey of the ex- jointly and, moreover, to consider two types
tended rough set approach to different MCDA of attributes: qualitative (symbolic, nominal,
problems has been recently presented in Greco like location) and quantitative (numerical, like
et al. (2001a). number of personnel).
In this paper, we wish to concentrate on the No other known approach to MCDA has all
extended rough set approach to one of major the above features. Our idea of extension of RST
classes of MCDA problems, which is the multiple- to deal with MA&C sorting problems consists in
criteria sorting problem. This problem is stated as building rough approximations of decision classes
follows: given a set of actions evaluated by a set of using a generalized ‘‘granule of knowledge’’. It is
criteria, assign these actions to some pre-defined defined by three relations considered jointly: in-
and preference-ordered decision classes, such that discernibility, defined on qualitative attributes,
each action is assigned to exactly one class. The similarity, defined on quantitative attributes, and
sorting problem is very similar to the problem of dominance, defined on criteria. The indiscernibility
classification; however, in the latter, the actions are relation is reflexive, symmetric and transitive,
evaluated by regular attributes and the decision while the similarity relation is reflexive and the
classes are not necessarily preference-ordered. The dominance relation is reflexive and transitive.
novelty of our present study is a joint consider- Decision rules with a special syntax are further
ation of criteria and regular attributes in one induced from these approximations of decision
sorting problem. Such a problem will be called an classes. The above idea builds on our previous
MA&C sorting problem. Although it has many study of MA&C sorting problems (Greco et al.,
counterparts in real life, it has not been considered 1998b), where only dominance and indiscernibility
either within MCDA or within methodologies relations were used together in the rough set ap-
dealing with multiple-attribute classification proach.
problems, like discriminant analysis, machine In Section 2, we define multicriteria and mul-
learning or neural networks. tiattribute rough approximations. Then, in Section
The interest of using the extended RST to 3, we characterize the syntax and semantics of
MA&C sorting problems is multiple indeed: decision rules induced from rough approxima-
1. It permits one to build the preference model of tions. Section 4 presents an illustrative example
a DM from a set of decision examples (s)he has solved by the original RST and by the extended
provided; these are exemplary assignments of RST. Section 5 contains conclusions.
some reference actions to the decision classes.
2. The preference model is in the form of a finite
set of specific decision rules, which ensures in- 2. Multicriteria and multiattribute rough approxi-
telligibility of the model; it also enables an easy mation
definition of possible strategies for passing from
one class to a better class through improvement For algorithmic reasons, knowledge about ref-
on some criteria. erence actions is represented in the form of an in-
3. When building the preference model, it does not formation table. The rows of the table are labelled
eliminate inconsistent examples that result from by reference actions, whereas columns are labelled
250 S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259

by attributes and entries of the table are attribute relation, i.e. reflexive, symmetric and transitive;
values, called descriptors. Formally, by an infor- it concerns mainly qualitative attributes;
mation table we understand the 4-tuple 2. similarity relation Rq , which is only reflexive; it
S ¼ hU ; Q; V ; f i, where U is a finite set of reference
S concerns mainly quantitative attributes (see
actions, Q is a finite set of attributes, V ¼ q2Q Vq Slowinski and Vanderpooten, 1997, 2000).
and Vq is a domain of the attribute q, and f : U  We introduce the following notation for disjoint
Q ! V is a total function such that f ðx; qÞ 2 Vq for subsets of C: C > is a subset of criteria, C ¼ is a
every q 2 Q, x 2 U , called an information function subset of qualitative attributes and C is a subset of
(see Pawlak, 1991). quantitative attributes. Therefore, C > [ C ¼ [
An information table can be seen as a decision C ¼ C and C > \ C ¼ ¼ ;, C > \ C ¼ ; and C ¼ \
table if in the set of attributes Q there are distin- C ¼ ;. Moreover, for any P  C we denote by P > ,
guished two disjoint sets: set C of condition attri- P ¼ and P the following subsets of P , respectively:
butes and set D of decision attributes. • the subset of criteria for which outranking rela-
As was mentioned in Section 1, the notion of tions are considered, i.e. P > ¼ P \ C > ,
attribute differs from that of criterion because the • the subset of qualitative attributes for which in-
domain (scale) of a criterion has to be ordered discernibility relations are considered, i.e. P ¼ ¼
according to a decreasing or increasing preference, P \ C¼ ,
while the domain of the attribute does not have to • the subset of quantitative attributes for which
be ordered. Formally, for each q 2 C being a cri- similarity relations are considered, i.e. P ¼
terion there exists an outranking relation (Roy, P \C .
1985) Sq on the set of actions U such that xSq y For each x; y 2 U we say that action x domi-
means ‘‘x is at least as good as y with respect to nates action y on criteria from P > if and only if
criterion q’’. We suppose that Sq is a total pre-or- xSq y for all q 2 P > . The dominance relation is re-
der, i.e. a strongly complete and transitive binary flexive and transitive. Similarly, we say that x is
relation defined on U on the basis of evaluations indiscernible from y on attributes from P ¼ if and
f ð ; qÞ. If domain Vq of criterion q is quantitative only if xIq y for all q  P ¼ , and x is similar to y on
and for each x; y 2 U , f ðx; qÞ P f ðy; qÞ implies attributes from P if and only if xRq y for all
xSq y, then Vq is a scale of preference of criterion q. qP .
If, however, for criterion q, Vq is not quantitative, Furthermore, assuming that the set of decision
or f ðx; qÞ P f ðy; qÞ does not imply xSq y, then in attributes D (possibly a singleton fdg) makes a
order to define a scale of preference of criterion q partition of the set of actions U into a finite number
one can choose a function gq : U ! R such that of decision classes, let Cl ¼ fClt ; t 2 T g,
for each x; y 2 U , xSq y if and only if gq ðxÞ P gq ðyÞ T ¼ f1; . . . ; ng, be a set of these classes such that
(see, e.g., Roubens and Vincke, 1985); passing each x 2 U belongs to one and only one Clt 2 Cl.
from f ðx; qÞ to gq ðxÞ for every x 2 U is called re- We suppose that the classes are ordered, i.e. for all
coding of the domain Vq . To recode Vq it is enough r; s 2 T , such that r > s, the actions from Clr are
to order the actions of U from the worst to the best preferred (strictly or weakly, Roy, 1985) to the ac-
on criterion q and to assign to gq ðxÞ consecutive tions from Cls . More formally, if S is a compre-
numbers corresponding to the rank of x in this hensive outranking relation on U , i.e. if for all
order, i.e. for z being the worst, gq ðzÞ ¼ 1, for w x; y 2 U , xSy means ‘‘x is at least as good as y’’, we
being the second worst, gq ðwÞ ¼ 2, and so on. suppose: ½x 2 Clr ; y 2 Cls ; r > s ) ½xSy and not
Then, the domain of function gq ð Þ becomes the ySx. The above assumptions are typical for con-
scale of preference of criterion q. sideration of a multiple-criteria sorting problem.
For each attribute q 2 C which is not a crite- In other words, the DM has assigned the ref-
rion, we admit a possibility of defining one of the erence actions to the classes of Cl according to the
two following binary relations on U : following comprehensive evaluation: the worst
1. indiscernibility relation Iq , which, according to actions are in Cl1 , the best actions are in Cln , the
the original rough sets theory, is an equivalence other actions belong to the remaining classes Clr ,
S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259 251

according to the principle that the higher r 2 T , Given P  C and x 2 U , the ‘‘granules of knowl-
the better class Clr . For example, in the risk as- edge’’ used for approximation in the extended
sessment problem, a financial manager can con- approach are:
sider Cl ¼ fCl1 ; Cl2 ; Cl3 g, where Cl1 is a class of • a set of actions y dominating action x on P > and
firms with ‘‘high risk’’, Cl2 is a class of firms with indiscernible from x on P ¼ , and such that x is
‘‘medium risk’’ and Cl3 is a class of firms with similar to y on P , DLþ P ðxÞ ¼ fy 2 U : yDP xg,
‘‘low risk’’. • a set of actions y dominating action x on P > and
The sets to be approximated are called the up- indiscernible from x on P ¼ , and such that y is

ward union and downward union of decision classes, similar to x on P , DUþ P ðxÞ ¼ fy 2 U : yDP xg,
respectively: • a set of actions y dominated by action x on P >
[ [ and indiscernible from x on P ¼ , and such that
CltP ¼ Cls ; Clt6 ¼ Cls ; t ¼ 1; . . . ; n: 
sPt s6t
x is similar to y on P , DL
P ðxÞ ¼ fy 2 U : xDP yg,
• a set of actions y dominated by action x on P >
The statement x 2 CltP means ‘‘x belongs at least and indiscernible from x on P ¼ , and such that
to class Clt ’’, while x 2 Clt6 means ‘‘x belongs at y is similar to x on P ; DU
P ðxÞ ¼ fy 2 U : xDP yg.
most to class Clt ’’. Let us remark that Cl1P ¼ The sets DLþ ðxÞ Uþ
P and D P ðxÞ are called ‘‘in-sim-
Cln6 ¼ U , ClnP ¼ Cln and Cl16 ¼ Cl1 . Further- dominating’’ sets, and the sets DL U
P ðxÞ and DP ðxÞ,
more, for t ¼ 2; . . . ; n we have ‘‘in-sim-dominated’’ sets.
6
Clt1 ¼ U  CltP and CltP ¼ U  Clt1
6
: For any P  C we say that x 2 U belongs to
CltP without any ambiguity if x 2 CltP and for each
The key idea of the rough set philosophy is action y 2 U dominating x on P > and indiscernible
approximation of one knowledge by another from x on P ¼ , and such that x is similar to y on P ,
knowledge. In the original approach, the knowl- also y belongs to CltP , i.e. DLþ P
P ðxÞ  Clt . Fur-
edge being approximated is a partition of U into thermore, we say that y 2 U could belong to CltP if
decision classes generated by a set of decision at- there exists at least one action x 2 CltP such that y
tributes; the knowledge used for approximation is dominates x on P > , y is indiscernible from x on P ¼
a partition of U into elementary sets of actions and y is similar to x on P , i.e. y 2 DUþP ðxÞ.
that are indiscernible by a set of condition attri- Thus, with respect to P  C, the set of all ac-
butes. The elementary sets are seen as ‘‘granules of tions belonging to CltP without any ambiguity
knowledge’’ used for approximation. constitutes the P -lower approximation of CltP ,
In our case, where among condition attributes denoted by P ðCltP Þ, and the set of all actions that
there are criteria, and decision classes are prefer- could belong to CltP constitutes the P -upper ap-
ence-ordered, the knowledge being approximated proximation of CltP , denoted by P ðCltP Þ:
is a collection of upward and downward unions of
P ðCltP Þ ¼ fx 2 U : DLþ P
P ðxÞ  Clt g;
decision classes and the ‘‘granules of knowledge’’ [
are sets of actions defined using dominance, simi- P ðCltP Þ ¼ DUþ
P ðxÞ; for t ¼ 1; . . . ; n:
larity and indiscernibility relations together, in- x2CltP
stead of indiscernibility relation alone. This is the
main difference between the original rough set Analogously, using DL U
P ðxÞ and DP ðxÞ, one can

approach and the present extension. define P -lower approximation and P -upper ap-
For any P  C let us define two reflexive binary proximation of Clt6 :
relations on U , denoted by DP and DP , and called P ðClt6 Þ ¼ fx 2 U : DL 6
P ðxÞ  Clt g;
‘‘in-sim-dominance’’ relations, such that for each [
x; y 2 U : P ðClt6 Þ ¼ DU
P ðxÞ; for t ¼ 1; . . . ; n:
• xDP y if and only if xSq y for each q 2 P > , xIq y for x2Clt6

each q 2 P ¼ and yRq x for each q 2 P and


• xDP y if and only if xSq y for each q 2 P > , xIq y for Theorem 1 (Rough inclusion). For any t 2 T and
each q 2 P ¼ and xRq y for each q 2 P . for any P  C,
252 S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259

P ðCltP Þ  CltP  P ðCltP Þ; P ðCltP Þ  RðCltP Þ; P ðClt6 Þ  RðClt6 Þ;

P ðClt6 Þ  Clt6  P ðClt6 Þ: P ðCltP Þ  RðCltP Þ; P ðClt6 Þ  RðClt6 Þ;


BnP ðCltP Þ  BnR ðCltP Þ; BnP ðClt6 Þ  BnR ðClt6 Þ:
Proof. Let x 2 P ðCltP Þ. DP being reflexive, x 2
DLþ Lþ P
P ðxÞ. Since DP ðxÞ  Clt , we have x 2 Clt .
P
P  Uþ Proof. From the definition of the ‘‘in-sim-domi-
Let x 2 Clt . DP being reflexive, x 2 DP ðxÞ, which
implies x 2 P ðCltP Þ. Analogous proof holds for nance’’ relations it follows that for any P  R  C,
P ðClt6 Þ  Clt6  P ðClt6 Þ.  xDR y implies xDP y and xDR y implies xDP y, for each
x; y 2 U . From the first implication we have
Theorem 2 (Complementarity). For any P  C, DLþ Lþ
P ðxÞ  DR ðxÞ for each x 2 U and, in conse-
quence, DP ðxÞ  CltP implies DLþ

R ðxÞ  Clt .
P
P ðCltP Þ ¼ U  P ðClt1
6
Þ; t ¼ 2; . . . ; n;
From the second implication it follows that
P
P ðClt6 Þ ¼ U  P ðCltþ1 Þ t ¼ 1; . . . ; n  1: DUþ Uþ
P ðxÞ  DR ðxÞ for each x 2 U . Thus, on the
basis of the definition of approximations and
Proof. If x 2 P ðCltP Þ, then DLþ P boundary of CltP , we obtain the proof with re-
P ðxÞ  Clt and,
therefore, there is no y 62 CltP such that yDP x. spect to CltP . Analogous proof holds for Clt6 . 
Since the complement of CltP in U is Clt16
, we can
P For any t 2 T and for any P  C, we define the
also say that x 2 P ðClt Þ if and only if there is no
6 6 accuracy of the approximation of CltP and Clt6 by
y 2 Clt1 such that yDP x. Since x 2 P ðClt1 Þ if and
P as the respective ratios
only if there exists at least one y 2 U such that
6
y 2 Clt1 and yDP x, this means that x 2 P ðCltP Þ if cardðP ðCltP ÞÞ
6
and only if x 62 P ðClt1 Þ. Thus, we have proved aP ðCltP Þ ¼ ;
P 6
cardðP ðCltP ÞÞ
that P ðClt Þ ¼ U  P ðClt1 Þ. Analogous proof
6
holds for P ðClt Þ ¼ U  P ðCltþ1P
Þ.  cardðP ðClt6 ÞÞ
aP ðClt6 Þ ¼ :
cardðP ðClt6 ÞÞ
The P -boundaries (doubtful regions) of the
The coefficient
unions CltP and Clt6 are defined, respectively, as
!
BnP ðCltP Þ ¼ P ðCltP Þ  P ðCltP Þ; [
cP ðClÞ ¼ card U  BnP ðClt6 Þ
BnP ðClt6 Þ ¼ P ðClt6 Þ  P ðClt6 Þ; for t ¼ 1; . . . ; n: t2T
!!,
[
[ BnP ðCltP Þ cardðU Þ
Theorem 3 (Identity of boundaries). For any t2T
t 2 T  f1g and for any P  C,
is called the quality of approximation of partition Cl
BnP ðCltP Þ ¼ BnP ðClt1
6
Þ:
by the set of attributes and criteria P , or in short,
quality of approximation. It expresses the ratio of
Proof. From Theorem 2 we have all P -correctly sorted actions to all actions in the
BnP ðCltP Þ ¼ P ðCltP Þ  P ðCltP Þ table. Due to identity of boundaries stated by
Theorem 3, the formula of cP ðClÞ can be rewritten
6 6
¼ ½U  P ðClt1 Þ  ½U  P ðClt1 Þ as
6 6
¼ P ðClt1 Þ  P ðClt1 Þ  S 
6
6 card U  t2T BnP ðClt Þ
¼ BnP ðClt1 Þ:  cP ðClÞ ¼
cardðU Þ
 S P

Theorem 4 (Monotonicity). For any t 2 T and for card U  t2T BnP ðClt Þ
¼ :
any P  R  C, cardðU Þ
S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259 253

Due to monotonicity stated by Theorem 4, it is of qualitative attributes, and the last to similarity
possible to define the reduct and the core within on a subset of quantitative attributes.
the extended RST. Each minimal subset P  C The following three types of decision rules can
such that cP ðClÞ ¼ cC ðClÞ is called a reduct of Cl be considered:
and denoted by REDCl ðCÞ. Let us remark that a • D P -decision rules which are generated from the
decision table can have more than one reduct. The lower approximation P ðCltP Þ, having the fol-
intersection of all reducts is called the core and lowing form:
denoted by CORECl ðCÞ. The algorithm for gen-
eration of all reducts in the dominance-based ‘‘if f ðx; q1 Þ P rq1 and . . . f ðx; qu Þ P rqu and
rough set approach has been presented in Susmaga f ðx; quþ1 Þ ¼ rquþ1 and . . . f ðx; qv Þ ¼ rqv and
et al. (1999). f ðx; qvþ1 Þ is similar to rqvþ1 and . . . f ðx; qz Þ is
similar to rqz , then x 2 CltP ’’, (STAT 1)

3. Decision rules where


P ¼ fq1 ; . . . ; qz g  C; P > ¼ fq1 ; . . . ; qu g;
The rough approximations of upward and
P ¼ ¼ fquþ1 ; . . . ; qv g and P ¼ fqvþ1 ; . . . ; qz g;
downward unions of classes can serve to induce a
generalized description of actions contained in the ðrq1 ; . . . ; rqz Þ 2 Vq1  Vq2   Vqz and t 2 T ;
information table in terms of ‘‘if . . . , then . . . ’’ • D 6 -decision rules which are generated from the
decision rules. For a given upward union of classes lower approximation P ðClt6 Þ, having the fol-
CltP the decision rules induced under a hypothesis lowing form:
that actions belonging to P ðCltP Þ are positive and
all the others negative suggest an assignment to ‘‘at ‘‘if f ðx; q1 Þ 6 rq1 and . . . f ðx; qu Þ 6 rqu and
least class Clt ’’. Analogously, for a given down- f ðx; quþ1 Þ ¼ rquþ1 and . . . f ðx; qv Þ ¼ rqv and
ward union Cls6 , the rules induced under a hy- f ðx; qvþ1 Þ is similar to rqvþ1 and . . . f ðx; qz Þ is
pothesis that actions belonging to P ðCls6 Þ are similar to rqz , then x 2 Clt6 ’’, (STAT 2)
positive and all the others negative suggest an as-
signment to ‘‘at most class Cls ’’. On the other where
hand, the decision rules induced under a hypoth-
esis that actions belonging to the intersection P ¼ fq1 ; . . . ; qz g  C; P > ¼ fq1 ; . . . ; qu g;
P ðCls6 Þ \ P ðCltP Þ are positive and all the others P ¼ ¼ fquþ1 ; . . . ; qv g and P ¼ fqvþ1 ; . . . ; qz g;
negative suggest an assignment to some class be-
ðrq1 ; . . . ; rqz Þ 2 Vq1  Vq2   Vqz and t 2 T ;
tween Cls and Clt ðs < tÞ.
We assume that domain Vq of each criterion • D P 6 -decision rules which are generated from
q 2 P > is its preference scale (eventually after re- the upper approximations P ðCls6 Þ \ P ðCltP Þ,
coding). Moreover, for each quantitative attribute having the following form:
q 2 P , we suppose that the similarity between
actions is represented by a similarity relation de- ‘‘if f ðx; q1 Þ P rq1 and . . . f ðx; qu Þ P rqu and
fined on the set of attribute values Vq . In other f ðx; quþ1 Þ 6 rquþ1 and . . . f ðx; qv Þ 6 rqv and
words, for each q 2 P and actions x; y 2 U , we f ðx; qvþ1 Þ ¼ rqvþ1 and . . . f ðx; qw Þ ¼ rqw and
conclude the similarity of x to y with respect to q f ðx; qwþ1 Þ is similar to rqwþ1 and . . . f ðx; qz Þ is
on the basis of the similarity of value f ðx; qÞ ¼ r to similar to rqz , then x 2 Cls [ Clsþ1 [
value f ðy; qÞ ¼ s, r; s 2 Vq . In the decision rules [Clt ’’, (STAT 3)
presented below, we use the similarity relation
among values. where
Each rule has three parts in the premise. The
first one corresponds to dominance on a subset of P ¼ fq1 ; . . . ; qz g  C;
criteria, the second to indiscernibility on a subset P > ¼ fq1 ; . . . ; qu g [ fquþ1 ; . . . ; qv g;
254 S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259

P ¼ ¼ fqvþ1 ; . . . ; qw g and P ¼ fqwþ1 ; . . . ; qz g; qwþ1 Þ ¼ rqwþ1 and . . . f ðy1 ; qz Þ ¼ rqz or f ðy2 ;


qz Þ ¼ rqz , and
ðrq1 ; . . . ; rqz Þ 2 Vq1  Vq2   Vqz
6. there exist no counterexample, i.e. for all z 2 U
and s; t 2 T such that s < t; observe that not such that f ðz; q1 Þ 6 rq1 and . . . f ðz; qu Þ P rqu and
necessarily fq1 ; . . . ; qu g \ fquþ1 ; . . . ; qv g ¼ ;: f ðz; quþ1 Þ 6 rquþ1 and . . . f ðz; qv Þ 6 rqv and f ðz;
quþ1 Þ ¼ rquþ1 and . . . f ðz; qv Þ ¼ rqv and f ðz; qvþ1 Þ
The statement (STAT 1) is accepted as a D P - is similar to rqvþ1 and . . . f ðz; qz Þ is similar to
decision rule if and only if rqz , we have z 2 Cls2 [ Cls2þ1 [ [ Clt2 ,
1. there exists at least one prototype y 2 P ðCltP Þ, s 6 s2 6 t2 6 t.
having the description on P ¼ fq1 ; . . . ; qz g Let us remark that conditions (2), (4) and (6)
equal to the right-hand sides (RHS) of elemen- above follow from asymmetry of the similarity
tary conditions in (STAT 1), i.e. f ðy; q1 Þ ¼ rq1 relation and its specific use in the definition of the
and . . . f ðy; qu Þ ¼ rqu and f ðy; quþ1 Þ ¼ rquþ1 and rough approximations. If the similarity relation
. . . f ðy; qv Þ ¼ rqv and f ðy; qvþ1 Þ ¼ rqvþ1 and were not considered, or if it were symmetric, then
. . . f ðy; qz Þ ¼ rqz , and conditions (1), (3) and (5) would be enough. Let us
2. there exist no counterexample, i.e. for all z 2 U explain this point in detail. For the sake of sim-
such that f ðz; q1 Þ P rq1 and . . . f ðz; qu Þ P rqu plicity we shall consider decision rules involving
and f ðz; quþ1 Þ ¼ rquþ1 and . . . f ðz; qv Þ ¼ rqv and conditions on similarity only. Consider a universe
f ðz; qvþ1 Þ is similar to rqvþ1 and . . . f ðz; qz Þ is sim- U ¼ fy; w; zg, set X ¼ fy; wg and an asymmetric
ilar to rqz , we have z 2 CltP . similarity relation defined as follows: y is similar to
w, z is similar to y (of course, each action from U is
Analogously, the statement (STAT 2) is ac-
similar to itself). The lower approximation of X is
cepted as a D 6 -decision rule if and only if
composed of y and w. Therefore, y could be chosen
3. there exists at least one prototype y 2 P ðClt6 Þ
as a prototype for a rule: ‘‘if x is similar to y, then x
having the description on P ¼ fq1 ; . . . ; qz g
belongs to X ’’. However, applying this rule to
equal to RHS of elementary conditions in
universe U , one could assign action z to X , which
(STAT 2), i.e. f ðy; q1 Þ ¼ rq1 and . . . f ðy; qu Þ ¼
is not true. Thus, to avoid such an anomaly, apart
rqu and f ðy; quþ1 Þ ¼ rquþ1 and . . . f ðy; qv Þ ¼
from the condition that the prototype action for a
rqv and f ðy; qvþ1 Þ ¼ rqvþ1 and . . . f ðy; qz Þ ¼ rqz ,
rule must belong to the lower approximation of X
and
[conditions (1), (3) and (5)], we impose the condi-
4. there exist no counterexample, i.e. for all z 2 U
tion that there is no action from outside X that is
such that f ðz; q1 Þ 6 rq1 and . . . f ðz; qu Þ 6 rqu and
similar to the prototype (conditions (2), (4) and
f ðz; quþ1 Þ ¼ rquþ1 and . . . f ðz; qv Þ ¼ rqv and f ðz;
(6)).
qvþ1 Þ is similar to rqvþ1 and . . . f ðz; qz Þ is similar
Since each decision rule is an implication, by a
to rqz , we have z 2 Clt6 :
minimal decision rule we understand such an im-
Finally, the statement (STAT 3) is accepted as a plication that there is no other implication with an
D P 6 -decision rule if and only if antecedent of at least the same weakness and a
5. there exist at least two prototypes consequent of at least the same strength. More
y1 ; y2 2 CðCls61 Þ \ CðCltP 1
Þ, s 6 s1 6 t1 6 t, not formal definition of minimal decision rules is pre-
necessarily different and not belonging to sented below.
CðCla6 Þ \ CðClbP Þ where a < s and/or b > t, A D P -decision rule ‘‘if f ðx; q1 Þ P rq1 and . . .
such that their description on P ¼ fq1 ; . . . ; qz g f ðx; qu Þ P rqu andf ðx; quþ1 Þ ¼ rquþ1 and . . . f ðx; qv Þ ¼
gives values to RHS of elementary conditions rqv and f ðx; qvþ1 Þ is similar to rqvþ1 and . . . f ðx; qz Þ is
in (STAT 3), i.e. y1 DP y2 and/or y1 DP y2 and similar to rqz , then x 2 CltP 1
’’ is minimal iff there is
f ðy2 ; q1 Þ ¼ rq1 and . . . f ðy2 ; qu Þ ¼ rqu and f ðy1 ; no other rule ‘‘if f ðx; p1 Þ P rp1 and . . . f ðx; pa ÞP rpa
quþ1 Þ ¼ rquþ1 and . . . f ðy1 ; qv Þ ¼ rqv and f ðy1 ; and f ðx; paþ1 Þ ¼ rpaþ1 and . . . f ðx; pb Þ ¼ rpb and
qvþ1 Þ ¼ f ðy2 ; qvþ1 Þ ¼ rqvþ1 and . . . f ðy1 ; qw Þ ¼ f ðx; pbþ1 Þ is similar to rpbþ1 and . . . f ðx; pc Þ is similar
f ðy2 ; qw Þ ¼ rqw and f ðy1 ; qwþ1 Þ ¼ rqwþ1 or f ðy2 ; to rpc , then x 2 CltP 2
’’ such that fp1 ; . . . ;
S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259 255

pa ; paþ1 ; . . . ; pb ; pbþ1 ; . . . ; pc g  fq1 ; . . . ; qu ; quþ1 ; . . . ; multiple-criteria sorting problem, a set of decision


qv ; qvþ1 ; . . . ; qz g, rj 6 rj for any j 2 fp1 ; . . . ; pa g and rules is complete if:
t2 P t1 . 1. Each action from lower approximation of any
A D 6 -decision rule ‘‘if f ðx; q1 Þ 6 rq1 and . . . upward or downward union of decision classes
f ðx; qu Þ 6 rqu and f ðx; quþ1 Þ ¼ rquþ1 and . . . f ðx; qv Þ ¼ is reassigned to the corresponding union by at
rqv and f ðx; qvþ1 Þ is similar to rqvþ1 and . . . f ðx; qz Þ is least one D P -decision rule or D 6 -decision rule
similar to rqz , then x 2 Clt61 ’’ is minimal iff there is from this set.
no other rule ‘‘if f ðx; p1 Þ 6 rq1 and . . . f ðx; pa Þ 6 rpa 2. Each pair of inconsistent actions belonging, re-
and f ðx; paþ1 Þ ¼ rpaþ1 and . . . f ðx; pb Þ ¼ rpb and spectively, to classes Cls and Clt ðs < tÞ is reas-
f ðx; pbþ1 Þ is similar to rpbþ1 and . . . f ðx; pc Þ is similar signed to a union of contiguous decision
to rpc , then x 2 Clt62 ’’ such that fp1 ; . . . ; pa ; classes, ranging from Cls to Clt , by at least
paþ1 ; . . . ; > pb ; pbþ1 ; . . . ; pc g  fq1 ; . . . ; qu ; quþ1 ; . . . ; one D P 6 -decision rule from this set.
qv ; qvþ1 ; . . . ; qz g and rj P rj for any j 2 fp1 ; . . . ; pa g Due to asymmetry of the similarity relation
and t2 6 t1 . considered in the definition of rough approxima-
A D P 6 -decision rule ‘‘if f ðx; q1 Þ P rq1 and . . . tions and due to its impact on the definition of
f ðx; qu Þ P rqu and f ðx; quþ1 Þ 6 rquþ1 and . . . f ðx; qv Þ decision rules, the above conditions of complete-
6 rqv and f ðx; qvþ1 Þ ¼ rqvþ1 and . . . f ðx; qw Þ ¼ rqw ness in the context of MA&C sorting problem
and f ðx; qwþ1 Þ is similar to rqwþ1 and . . . f ðx; qz Þ is have to be refined. Let us explain the reason for
similar to rqz , then x 2 Cls1 [ Cls1 þ1 [ [ Clt1 ’’ is this refinement in the following small example. For
minimal iff there is no other rule ‘‘if f ðx; p1 Þ P rp1 the sake of simplicity we shall consider decision
and . . . f ðx; pa Þ P rpa and f ðx; paþ1 Þ 6 rpaþ1 and . . . rules involving conditions of similarity only.
f ðx; pb Þ 6 rpb and f ðx; pbþ1 Þ ¼ rpbþ1 and . . . f ðx; Consider a universe U ¼ fv; y; w; zg, set
pc Þ ¼ rpc and f ðx; pcþ1 Þ is similar to rpcþ1 and . . . X ¼ fv; y; zg and an asymmetric similarity relation
f ðx; pd Þ is similar to rpd , then x 2 Cls2 [ Cls2 þ1 defined as follows: w is similar to y, z is similar to v
[ [ Clt2 ’’ such that fp1 ; . . . ; pa ; paþ1 ; . . . ; pb ; and to w (of course, each action from U is similar
pbþ1 ; . . . ; pc g  fq1 ; . . . ; qu ; quþ1 ; . . . ;qv ; qvþ1 ; . . . ; to itself). The lower approximation of X is com-
qz g, rj 6 rj for any j 2 fp1 ; . . . ; pa g, rj P rj for any posed of y and v; z does not belong to the lower
j 2 fpaþ1 ; . . . ; pb g, s2 P s1 and t2 6 t1 . approximation of X because it is also similar to w
An action y 2 U supports a decision rule if it from outside X . Although action y is in the lower
matches its condition part. Formally, y 2 U sup- approximation of X , it cannot be a prototype for a
ports the D P -decision rule (STAT 1) iff decision rule because it is not true that each action
f ðy; q1 Þ P rq1 and . . . f ðy; qu Þ P rqu and f ðy; quþ1 Þ ¼ similar to y belongs to X (w from outside X is
rquþ1 and . . . f ðy; qv Þ ¼ rqv and f ðy; qvþ1 Þ is similar similar to y). Therefore, there is no rule that could
to rqvþ1 and . . . f ðy; qz Þ is similar to rqz , and reassign univocally y to X , even if y belongs to
y 2 CltP . An action y 2 U supports the D 6 -deci- lower approximation of X .
sion rule (STAT 2) iff f ðy; q1 Þ 6 rq1 and . . . Remark, however, that action z will be reclas-
f ðy; qu Þ 6 rqu and f ðy; quþ1 Þ ¼ rquþ1 and . . . f ðy; sified to set X , even if it does not belong to the
qv Þ ¼ rqv and f ðy; qvþ1 Þ is similar to rqvþ1 and lower approximation of X . This reclassification is
. . . f ðy; qz Þ is similar to rqz , and y 2 Clt6 . Finally, possible due to the rule: ‘‘if x is similar to v, then x
an action y 2 U supports the D P 6 -decision rule belongs to X ’’, which is based on prototype v. Fi-
(STAT 3) iff f ðy; q1 Þ P rq1 and . . . f ðy; qu Þ P rqu and nally, y and w will be assigned (approximately) to X
f ðy; quþ1 Þ 6 rquþ1 and . . . f ðy; qv Þ 6 rqv and f ðy; qvþ1 Þ or to U  X by decision rule: ‘‘if x is similar to y,
is similar to rqvþ1 and . . . f ðy; qw Þ is similar to rqw , then x belongs to X or U  X , without enough in-
and y 2 Cls [ Clsþ1 [ [ Clt . formation to assign x exactly to one of these sets’’.
The number of actions supporting a rule is Reassignment of an action from U , matching
called its strength. some D P - and D 6 -decision rules, is made to a
Another property often required for a set of class (or a set of contiguous classes) resulting from
decision rules is completeness. In the case of the intersection of unions of decision classes suggested
256 S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259

by the rules. Moreover, reassignment of an action One of three induction strategies can be adop-
from U matching some D P 6 -decision rules is ted to obtain a set of decision rules (Stefanowski
made to a set of contiguous classes being a union and Vanderpooten, 1994; Mienko et al., 1996;
of classes suggested by the rules. Stefanowski, 1998; Slowinski et al., 2000):
In consequence of using the asymmetric simi- • generation of a minimal description, i.e. a mini-
larity relation, the above conditions of completeness mal set of rules,
should be read as: the set of actions reassigned by • generation of an exhaustive description, i.e. all
the complete set of rules to original decision classes possible minimal rules for a given information
should be maximal and the assignment should be table,
the most precise possible. Let us explain what • generation of a characteristic description, i.e. a
‘‘maximal’’ and ‘‘most precise’’ means. This means set of minimal rules covering relatively many ac-
that no other rule can be induced which would re- tions each, however, altogether not necessarily
assign more actions to original classes or to smaller all actions from U .
sets of contiguous classes. For example, a set of rules
is not complete if it assigns action x to classes Cl2
and Cl3 , but there exists another rule, not included 4. An example
in this set, permitting precise assignment of x to Cl3 .
Analogously, a set of rules is not complete if it as- The following example (extending the one given
signs action x to classes Cl2 , Cl3 and Cl4 , but there by Pawlak, 1997) illustrates the concepts introduced
exists another rule, not included in this set, permit- above and shows the main differences between the
ting more precise assignment of x to Cl3 and Cl4 . original RST and the present extension. In Table 2,
The maximal set of actions that can be reclas- there are shown exemplary decisions of a DM con-
sified exactly by a complete set of rules can be cerning eight warehouses (reference actions) de-
formally characterized as follows: scribed by means of three condition attributes:
1. For each P  C, the set LtP ðP Þ ¼ fw 2 U : • A1 , capacity of the sales staff,
DLþ P
and DUþ P • A2 , geographical region,
P ðwÞ  Clt P ðwÞ  Clt g is com-
posed of all prototypes of D P -decision rules in- • A3 , area.
volving criteria and attributes from P . The decision attribute d specifies the assignment
Therefore, one can reclassify by D P -decision made by the DM into two sets of warehouses
rules based on these prototypes all actions y making either profit or loss. Table 2 represents
such that there exists P  C and w 2 LtP ðP Þ preferential information given by the DM.
for which yDP w.
2. For each P  C, the set y 2 Lt6 ðP Þ ¼ 4.1. The results of the original rough set approach
fw 2 U : DL 6
P ðwÞ  Clt and DP
U
ðwÞ  Clt6 g
is composed of all prototypes of D 6 -decision Class Cl1 of the warehouses making loss and
rules involving criteria and attributes from P . class Cl2 of the warehouses making profit were
Therefore, one can reclassify by D 6 -decision
rules based on these prototypes all actions y
such that there exists P  C and w 2 Lt6 ðP Þ Table 2
Decision table with exemplary decisions
for which wDP y.
All actions that cannot be reclassified exactly by Warehouse A1 A2 A3 d
D P - and D 6 -decision rules are assigned to sets of w1 Medium A 500 Loss
w2 Good A 400 Profit
contiguous classes by D P 6 -decision rules, so that
w3 Medium A 450 Profit
all actions are classified by a complete set of rules. w4 Good B 400 Loss
We call minimal each set of minimal decision w5 Good B 475 Profit
rules that is complete and non-redundant, i.e. ex- w6 Medium B 425 Profit
clusion of any rule from this set makes it non- w7 Medium B 350 Profit
w8 Medium B 350 Loss
complete.
S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259 257

approximated by means of the original rough set C ¼ fA1 ; A2 ; A3 g and D ¼ fdg; however, we must
approach. It is clear that C ¼ fA1 ; A2 ; A3 g and observe that A1 is a criterion, while A2 and A3 are
D ¼ fdg. The C-lower approximations, the C-up- attributes. With respect to A1 , ‘‘good’’ is better
per approximations and the C-boundaries of sets than ‘‘medium’’, while with respect to A2 and A3 no
Cl1 and Cl2 are, respectively: preference is expressed on their domains. More-
over, for attribute A2 , the classical indiscernibility
CðCl1 Þ ¼ fw1; w4g; relation is considered, i.e. for each x; y 2 U , xIq y if
and only if f ðx; A2 Þ ¼ f ðy; A2 Þ, and for attribute
CðCl1 Þ ¼ fw1; w4; w7; w8g;
A3 , a similarity relation is defined such that for
BnC ðCl1 Þ ¼ fw7; w8g; each x; y 2 U , xRq y if and only if

CðCl2 Þ ¼ fw2; w3; w5; w6g; j f ðy; A3 Þ  f ðx; A3 Þj


6 10%:
f ðy; A3 Þ
CðCl2 Þ ¼ fw2; w3; w5; w6; w7; w8g;
The C-lower approximations, the C-upper ap-
BnC ðCl2 Þ ¼ fw7; w8g: proximations and the C-boundaries of sets C16 and
Cl1P are, respectively:
Therefore, the accuracy of the approximation is
equal to 0.5 for the class of warehouses making CðCl16 Þ ¼ fw1g;
loss and to 0.67 for the class of warehouses making
profit, while the quality of classification is equal to CðCl16 Þ ¼ fw1; w3; w4; w6; w7; w8g;
0.75. There is only one reduct which is also the
BnC ðCl16 Þ ¼ fw3; w4; w6; w7; w8g;
core, i.e. REDCl ðCÞ ¼ CORECl ðCÞ ¼ fA2 ; A3 g.
Continuing the original rough set approach, the CðCl2P Þ ¼ fw2; w5g;
following minimal set of decision rules has been
induced from the considered decision table using CðCl2P Þ ¼ fw2; w3; w4; w5; w6; w7; w8g;
the LERS induction algorithm (Grzymala-Busse, BnC ðCl2P Þ ¼ fw3; w4; w6; w7; w8g:
1992) (within parentheses there are the actions
supporting the corresponding rules): Therefore, the accuracy of the approximation is
1. if f ðx; A3 Þ ¼ 500, then x 2 Cl1 , ðw1Þ equal to 0.17 for Cl16 and to 0.29 for Cl2P , while
2. if f ðx; A2 Þ ¼ B and f ðx; A3 Þ ¼ 400, then x 2 Cl1 , the quality of classification is equal to 0.37. There
ðw4Þ is only one reduct, which is also the core, i.e.
3. if f ðx; A2 Þ ¼ A and f ðx; A3 Þ ¼ 400, then x 2 Cl2 , REDCl ðCÞ ¼ CORECl ðCÞ ¼ fA2 ; A3 g.
ðw2Þ The following minimal set of decision rules can
4. if f ðx; A3 Þ ¼ 450, then x 2 Cl2 , ðw3Þ be obtained from the considered decision table
5. if f ðx; A3 Þ ¼ 475, then x 2 Cl2 , ðw5Þ (within parentheses there are the actions support-
6. if f ðx; A3 Þ ¼ 425, then x 2 Cl2 , ðw6Þ ing the corresponding rules and prototype actions
7. if f ðx; A3 Þ ¼ 350, then x 2 Cl1 or x 2 Cl2 . of D P - and D 6 -decision rules are marked in
ðw7; w8Þ bold):
1. if f ðx; A2 Þ ¼ A and f ðx; A3 Þ is similar to 500,
then x 2 Cl1 [ Cl2 , ðw1; w3Þ
4.2. The results of approximations by ‘‘in-sim- 2. if f ðx; A1 Þ is at least ‘‘good’’ and f ðx; A2 Þ ¼ A,
dominance’’ relations then x 2 Cl2P , ðw2Þ
3. if f ðx; A2 Þ ¼ B and f ðx; A3 Þ is similar to 400,
With this approach we approximate class Cl16 then x 2 Cl1 [ Cl2 , ðw4; w6Þ
of the warehouses making at most loss and the 4. if f ðx; A1 Þ is at least ‘‘good’’ and f ðx; A3 Þ is sim-
class Cl2P of the warehouses making at least profit. ilar to 475, then x 2 Cl2P , ðw5Þ
Since only two classes are considered, we have 5. if f ðx; A3 Þ is similar to 350, then
Cl16 ¼ Cl1 and Cl2P ¼ Cl2 . Also in this case x 2 Cl1 [ Cl2 . ðw7; w8Þ
258 S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259

4.3. Comparison of the results due to the following facts: w3 is similar to w1 on


attribute A3 , w1 ‘‘in-sim-dominates’’ w3, such that
The advantages of the rough set approach w1DC w3, and it is not possible to construct a D 6 -
based on ‘‘in-sim-dominance’’ relations over the decision rule from w1 as prototype because w3 is a
original rough set analysis based on the indis- counterexample for such a rule, i.e. w3 is domi-
cernibility relation can be summarized in the fol- nated by w1 on A1 , w3 is indiscernible from w1 on
lowing points. A2 and w3 is similar to w1; however, w3 belongs to
Let us observe that the quality of classification Cl2 (see condition 4 for existence of a D 6 -decision
using indiscernibility is equal to 0.75 since ware- rule in Section 3).
houses w7; w8 are indiscernible while belonging to Let us also remark that the minimal set of de-
different classes. The quality of classification using cision rules induced from the decision table by the
‘‘in-sim-dominance’’ is equal to 0.375 due to ad- new approach gives a smaller number of stronger
ditional inconsistency introduced by warehouses rules (5 against 7).
w3, w4, and w6. For example, with respect to cri- The minimal set of rules based on ‘‘in-sim-
terion A1 , w4 outranks w6; with respect to attribute dominance’’ represents a preference model of the
A2 , w4 and w6 are indiscernible; with respect to the DM that is perfectly compatible with the DM’s
attribute A3 , w6 is similar to w4. Therefore, w4 ‘‘in- exemplary decisions and, moreover, reveals all
sim-dominates’’ w6, such that w4DC w6 (also kinds of inconsistency (with respect to indiscern-
w4DC w6, since w4 is similar to w6). However, w4 ibility, similarity and dominance) in the preferen-
has a comprehensive evaluation worse (loss) than tial information. The reaction of the DM to such a
w6 (profit). Therefore, this can be interpreted as an model of preferences could be either its acceptance
inconsistency revealed by the approximation based or a revision of assignment examples.
on ‘‘in-sim-dominance’’ that cannot be captured
when the approximation is based on indiscern-
ibility and/or similarity only. 5. Conclusions
Let us remark that, according to what was
written at the end of Section 3, the reassignment of We presented a new rough set method for
the eight warehouses using the minimal set of rules MA&C sorting problems whose purpose is to ap-
based on ‘‘in-sim-dominance’’ gives the following proximate a partition of some reference actions
result (‘‘!’’ means assignment): described by criteria and by attributes, which are
not criteria, into pre-defined and preference-or-
w1 ! Cl1 or Cl2 ðloss or profitÞ; dered decision classes. The ‘‘granules’’ of knowl-
w2 ! Cl2 ðprofitÞ; edge with which the partition is (roughly)
approximated are defined using a dominance re-
w3 ! Cl1 or Cl2 ðloss or profitÞ; lation on criteria and indiscernibility or similarity
relation on attributes. We showed that the basic
w4 ! Cl1 or Cl2 ðloss or profitÞ; concepts of rough sets theory are restored in this
w5 ! Cl2 ðprofitÞ; new methodology. The comparison of the original
rough set approach with the extended rough set
w6 ! Cl1 or Cl2 ðloss or profitÞ; approach on a didactic example gives evidence of
advantages of the new methodology.
w7 ! Cl1 or Cl2 ðloss or profitÞ; Because of frequent heterogeneity in descrip-
w8 ! Cl1 or Cl2 ðloss or profitÞ: tion of decision situations, the MA&C sorting
problems are quite realistic but there was no
Although warehouse w1 belongs to C-lower methodology for dealing with them. The prefer-
approximation of Cl16 , it is not reassigned to Cl1 ence model is inferred from a set of assignment
but to Cl1 or Cl2 because of asymmetry of the examples provided by the DM. The preference
similarity relation on A3 . More precisely, this is model has an unusual form of ‘‘if . . . , then . . . ’’
S. Greco et al. / European Journal of Operational Research 138 (2002) 247–259 259

decision rules. They satisfy conditions of com- Greco, S., Matarazzo, B., Slowinski, R., 2001a. Rough sets
pleteness, and dominance, and manage with pos- theory for multicriteria decision analysis. European Journal
of Operational Research 129, 1–47.
sible inconsistencies in the set of examples. Greco, S., Matarazzo, B., Slowinski, R., 2001b. Conjoint
Moreover, such a preference representation is measurement and rough set approach for multicriteria
more understandable for the users because of its sorting problems in presence of ordinal criteria. In: Colorni,
natural syntax. For this reason, it facilitates an A., Paruccini, M., Roy, B. (Eds.), AMCDA – Aide
interactive construction of the DM’s preference Multicritere a la Decision (Multiple Criteria Decision
Aiding). EUR Report, Joint Research Centre, The Euro-
model. pean Commission, Ispra, pp. 114–141.
Other extensions of RST for dealing with Grzymala-Busse, J.W, 1992. LERS – a system for learning from
problems of MCDA are described in Greco et al. examples based on rough sets. In: Slowinski, R. (Ed.),
(1999, 2000, 2001a). Intelligent Decision Support. Handbook of Applications
and Advances of the Rough Sets Theory. Kluwer Academic
Publishers, Dordrecht, pp. 3–18.
Mienko, R., Stefanowski J., Toumi, K., Vanderpooten, D.,
Acknowledgements 1996. Discovery-oriented induction of decision rules. Cahier
du LAMSADE no. 141, Universite de Paris Dauphine, Paris.
Pawlak, Z., 1991. Rough Sets. Theoretical Aspects of Reason-
The research of the first two authors has been ing about Data. Kluwer Academic Publishers, Dordrecht.
supported by the Italian Ministry of University Pawlak, Z., 1997. Rough set approach to knowledge-based
and Scientific Research (MURST). The third au- decision support. European Journal of Operational Re-
thor wishes to acknowledge financial support from search 99, 48–57.
Roubens, M., Vincke, Ph., 1985. Preference Modelling. Spring-
the State Committee for Scientific Research, grant
er, Berlin.
KBN 8T11F 006 19. Roy, B., 1985. Methodologie Multicritere d’Aide a la Decision.
Economica, Paris.
Slowinski, R. (Ed.), 1992. Intelligent Decision Support. Hand-
book of Applications and Advances of the Rough Sets
References Theory. Kluwer Academic Publishers, Dordrecht.
Slowinski, R., Stefanowski, J., Greco, S., Matarazzo, B., 2000.
Greco, S., Matarazzo, B., Slowinski, R., 1996. Rough approx- Rough sets based processing of inconsistent information in
imation of a preference relation by dominance relations. decision analysis. Control and Cybernetics 29, 379–404.
ICS Research Report 16/96, Warsaw University of Tech- Slowinski, R., Vanderpooten, D., 1997. Similarity relation as a
nology, Warsaw. European Journal of Operational Re- basis for rough approximations. In: Wang, P.P. (Ed.),
search 117, 63–83 (1999). Advances in Machine Intelligence and Soft Computing.
Greco, S., Matarazzo, B., Slowinski, R., 1998a. A new rough Duke University Press, Durham, NC, pp. 17–33.
set approach to evaluation of bankruptcy risk. In: Zo- Slowinski, R., Vanderpooten, D., 2000. A generalised definition
pounidis, C. (Ed.), Operational Tools in the Management of of rough approximations. IEEE Transactions on Data and
Financial Risks. Kluwer Academic Publishers, Dordrecht, Knowledge Engineering 12, 331–336.
pp. 121–136. Stefanowski, J., 1998. On rough set based approaches to
Greco, S., Matarazzo, B., Slowinski, R., 1998b. A new rough induction of decision rules. In: Polkowski, L., Skowron, A.
set approach to multicriteria and multiattribute classifica- (Eds.), Rough Sets in Knowledge Discovery, vol. 1. Physica-
tion. In: Polkowski, L., Skowron, A. (Eds.), Rough Sets and Verlag, Heidelberg, pp. 500–529.
Current Trends in Computing, Lecture Notes in Artificial Stefanowski, J., Vanderpooten, D., 1994. A general two stage
Intelligence, vol. 1424. Springer, Berlin, pp. 60–67. approach to rule induction from examples. In: Ziarko, W.
Greco, S., Matarazzo, B., Slowinski, R., 1999. The use of rough (Ed.), Rough Sets, Fuzzy Sets and Knowledge Discovery.
sets and fuzzy sets in MCDM. In: Gal, T., Stewart, T., Springer, Berlin, pp. 317–325.
Hanne, T. (Eds.), Advances in Multiple Criteria Decision Susmaga, R., Slowinski, R., Greco, S., Matarazzo, B., 1999.
Making. Kluwer Academic Publishers, Dordrecht, pp. 14.1– Computation of reducts for multi-attribute and multi-
14.59 (Chapter 14). criteria classification. In: Proceedings of the VIIth Work-
Greco, S., Matarazzo, B., Slowinski, R., 2000. Extension of the shop on Intelligent Information Systems. Institute of
rough set approach to multicriteria decision support. Computer Science Fundamentals, Polish Academy of Sci-
INFOR Journal (Canadian Journal of Operational Re- ences, Warsaw, pp. 154–163.
search and Information Processing) 38, 161–196.

You might also like