Intelligent DSS Under Verbal Decision Analysis
Ilya Ashikhmin (Cork Constraint Computation Centre – University College Cork, Ireland), Eugenia Furems (Russian Academy of Sciences, Russia), Alexey Petrovsky (Institute for Systems Analysis – Russian Academy of Sciences, Russia) and Michael Sternin (Institute for Systems Analysis – Russian Academy of Sciences, Russia)
Copyright: © 2008
Verbal decision analysis (VDA) is a relatively new term introduced in Larichev and Moshkovich (1997) for a methodological approach to discrete multi-criteria decision making (MCDM) problems that was under elaboration by Russian researchers since the 1970s. Its main ideas, principles, and strength in comparison with other approaches to MCDM problems are summarized in Moshkovich, Mechitov, and Olson (2005) and in posthumous book (Larichev, 2006) as follows: problem description (alternatives, criteria, and alternatives’ estimates upon criteria) with natural language without any conversion to numerical form; usage of only those operations of eliciting information from a decision maker (DM) that deems to be psychologically reliable; control of DM’s judgments consistency, and traceability of results, that is, the intermediate and final results of a problem solution have to be explainable to DM. The main objective of this chapter is to provide an analysis of the methods and models of VDA for implementing them in intellectual decision support systems. We start with an overview of existing approaches to VDA methods and model representation. In the next three sections we present examples of implementing the methods and models of VDA for intellectual decision support systems designed for such problems solving as discrete multi-criteria choice, construction of expert knowledge base, and multi-criteria assignment problem. Finally, we analyze some perspective of VDA-based methods to implement them for intellectual decision support systems.
Key Terms in this Chapter
Verbal Decision Analysis (VDA): VDA is a methodological approach to discrete multi-criteria decision making (MCDM) problems. Its main ideas, principles, and strengths in comparison with other approaches to MCDM problems are: problem description (alternatives, criteria, and alternatives’ estimates upon criteria) with natural language; usage of only those operations of eliciting information from a decision maker (DM) that deems to be psychologically reliable without any conversion to numerical form; control of the DM’s judgments consistency; and traceability of results, that is, the intermediate and final results of a problem solving have to be explainable to the DM.
Multi-Criteria Assignment Problem (MAP): MAP is characterized with many parameters, criteria, and goals. Such peculiarities require taking into account individual preferences, quality criteria, uncertainty, and the large dimension.
Expert Knowledge Acquisition (EKA): EKA is for diagnostic decision support systems (DDSS) and may be stated as multi-attribute classification problem (MAC). In the context of EKA, MAC may be stated in nominal terms, where the expert decision rules based on his/her knowledge are used instead of his/her preferences.
Discrete Decision Making Problems (DDMP): DDMP is a generalized morphological scheme, which involves such typical multi-criteria/multi-attribute problems as Choice, Group Ranking, Linear Ranking, and Clustering, including both ordinal and nominal classification.
UniComBOS: UniComBOS is a method for multi-criteria comparison and choice and intelligent decision support system (IDSS) designed to assist a DM in choosing the best alternative from the finite set given their qualitative performance estimates upon multiple criteria.
Mutual Preference Independence: Criteria set C is mutually preference independent if for any subset C*? C the preference between any pair of alternatives differing only upon criteria C* does not depend on fixed equal estimates upon criteria C\C*.