Hybrid method of forming preferences function in computerized decision support systems

Аuthors

Eskin V. I.^1*, Sudakov V. A.^2**

1. ,
2. Keldysh Institute of Applied Mathematics (Russian Academy of Sciences), 4, Miusskaya sq., Moscow, 125047, Russia

*e-mail: kaf302@mai.ru
**e-mail: sudakov@keldysh.ru

Abstract

The main purpose of this work is the construction of the preference functions formation method that allows to take into account the dependence of preferences that does not require a significant investment of time to build a user preference functions, working with vector criterion of arbitrary dimension.
The original method for hybrid functions preferences formation is proposed. The user selects disjoint regions in the space of criteria. For each of these areas the user sets a preference value of the level in the scale of preferences, which set strict preference relation. Then a set of convex criteria space subregions is constructed in automatic mode according the next conditions: they do not overlap with the areas specified, and the union of all subdomains formed with source areas provides a complete criteria space.
Next, using the concept of Pareto dominance and a number of other methods of determining dominance we can define a preference relation between the areas specified by the user, and the areas identified automatically.
The next step is to build a directed graph whose vertices correspond to the regions and the edges are the preference relation for vertices. Next, the user needs to disassemble the graph similar to that suggested in Larichev ZAPROS method.
In some practical problems of ranking alternatives several non-dominated alternatives often fall into one area, so that a their relation must be defined. In this case the user should either divide this area into smaller ones, or use the formal (quantitative) method of comparing alternatives within a given area. A function to compute the total value of preferences is called a hybrid function of preferences (HFP).
Currently, we implement the interface for specified HFP method in C # and ASP.NET. Initial data are stored and processed in DSS database under control of MS SQL.
The above method of constructing hybrid preference functions allows one to take into account the dependence on the preferences, provides the required accuracy of decision support, contains a fast iterative procedure for entering and editing preferences. The method is invariant with respect to the subject area, it is planned to use in tasks ranging applications for scientific and applied research on the Russian segment of the International Space Station.

Keywords:

relationship of domination, decision support system, vector criterion, qualitative and quantitative methods for decision support, a hybrid function of preferences

References

1. Osipov V.P., Sudakov V.A., Khakhulin G.F. Vestnik kompyuternykh iinformatsionnykh tekhnologii, 2012, no.12, pp. 24-28.
2. Afonin K.A., Bomas V.V., Eskin V.I. Metody isistemy prinyatiya reshenii. Avtomatizatsiya iintellektualizatsiya protsessov proektirovaniya iupravleniya, Riga, Rizhskii tekhnicheskii universitet, 1993, pp. 23-25.
3. Bomas V.V., Sudakov V.A. Podderzhka sub^,,ektivnykh reshenii vmnogokriterialnykh zadachakh (Support for subjective decisions inamulticriterial tasks), Moscow, MAI, 2011, 176p.
4. Podinovskii V.V. Vvedenie vteoriyu vazhnosti kriteriev vmnogokriterialnykh zadachakh prinyatiya reshenii (Introduction toimportance ofcriteria inmulticriterial decision-making problems), Moscow, Fizmatlit, 2007, 64p.
5. Larichev O.I. Teoriya imetody prinyatiya reshenii, atakzhe Khronika sobytii vVolshebnykh stranakh (Theory and methods ofdecision-making and Chronicle tothe Magic country), Moscow, Logos, 2000, 296p.
6. Lotov A.V., Pospelova I.I. Mnogokriterialnye zadachi prinyatiya reshenii (Multicriterial decision making problems), Moscow, MAKS Press, 2008, 197p.