[Apdio] Seminário em Engenharia e Gestão do CEG-IST, 06/12/2007 (divulgação)

[<M.D.Oliveira@lse.ac.uk>]

Caros colegas,

 

O próximo Seminário em Engenharia e Gestão do CEG-IST realizar-se-à na próxima semana no dia 6 de Dezembro (quinta-feira) e tem a seguinte referência:

 

Data: 6 de Dezembro de 2007, 11.00-12.00
Localização, sala: Campus IST Taguspark, Anfiteatro A5
Apresentador: Salvatore Greco (Universidade da Catánia, Itália) (salgreco@unict.it <mailto:salgreco@unict.it> )
Título do seminário: " Dominance-based Rough Set Approach for Decision Analysis - a Tutorial " 

Summary: Scientific analysis of decision problems aims at giving the decision maker (DM) a recommendation concerning a set of objects (called also alternatives, solutions, acts, actions, cases, candidates) evaluated from the point of view of a plurality of their characteristics considered relevant for the problem at hand, and called attributes. The following three most frequent decision problems are typically distinguished:

*        classification, when the decision aims at assigning objects to predefined classes,

*        choice, when the decision aims at selecting the best objects,

*        ranking, when the decision aims at ordering objects from the best to the worst.

The above categorization can be refined with respect to classification Problems by distinguishing between:

*        nominal classification, called also taxonomy, when neither the value sets of attributes nor the predefined classes are preference ordered, 

*        ordinal classification, called also sorting,  when both the value sets of attributes and the predefined classes are preference ordered (even if in this case it is also possible to take into account attributes with non ordered value sets. 

If value sets of attributes are ordered according to preference, they are Called criteria, otherwise they keep the name of attributes. For scientific support of decisions one needs a more or less explicit model of the decision problem. The model relates the decision to the characteristics of the objects expressed by the considered attributes. Building such a model requires information about conditions and parameters of aggregation of multi-attribute characteristics of objects. Very often this information is not easily definable. Even if the required information is easily definable, it is often processed in a way that is not clear for the DM, such that he cannot see what are the exact relations between the provided information and the final recommendation. Consequently, very often the decision model is perceived by the DM as a black box whose result has to be accepted because the analyst's authority guarantees that the result is "right". In this context, the aspiration of the DM to find good reasons to make decision is frustrated and it rises the need for a more transparent methodology in which the relation between the original information and the final recommendation is clearly shown. Such a transparent methodology searched for can be called glass box. Its typical representative is based on application of decision rules which express the decision model in terms of a set of "if..., then..." statements. From one side, the decision rules are explicitly related to the original information and, from the other side, they give understandable justifications for the decision to be made. The rules induced from the input information provided in terms of exemplary decisions represent a transparent decision model for the DM, and enable his understanding of the reasons of his previous decisions. The acceptance of the rules by the DM justifies, in turn, their use for decision support. The induction of rules from examples is a typical approach of artificial intelligence. This explains our interest in rough set theory which proved to be a useful tool for analysis of vague description of decision situations. The aim of rough set analysis is the explanation of the dependence between the values of some decision attributes, playing the role of "dependent variables", by means of the values of other condition attributes, playing the role of "independent variables". An important advantage of the rough set approach is that it can deal with partly inconsistent data. Moreover, it provides useful information about the role of particular attributes and their subsets, and prepares the ground for representation of knowledge hidden in the data by means of  "if ..., then ..." decision rules. Classical Rough Set Approach (CRSA) proposed by Pawlak cannot deal with preference order in the value sets of condition and decision attributes. Thus, from among all listed above decision problems, classical rough set approach supports nominal classification only. For ordinal classification, choice and ranking it is necessary to generalize the classical rough set approach, so as to take into account preference orders. This generalization, called Dominance-based Rough Set Approach (DRSA), has been proposed by Greco, Matarazzo and Slowinski. Moreover, it has been proved that CRSA is a specific case of DRSA, and, therefore, any application of rough set approach to decision problems can be effectively dealt with using DRSA. In this tutorial we present applications of DRSA to some real life decision problems, emphasizing the advantages of this approach with respect to competitive approaches.

 

Os seminários do CEG-IST são de acesso livre.

 

Com os meus melhores cumprimentos,

 

Mónica Oliveira

CEG-IST @ https://fenix.ist.utl.pt/investigacao/CEG-IST <https://fenix.ist.utl.pt/investigacao/CEG-IST>  


Please access the attached hyperlink for an important electronic communications disclaimer: http://www.lse.ac.uk/collections/secretariat/legal/disclaimer.htm