Groshev

Quite often in real situations, the quality of operation of the object or system being examined is not estimated by a single criterion or quality measure, but by a set of such criteria that are equally significant. Tasks of this type relate to, and are an important stage in the design and evaluation of the design object. There are a large number of methods that allow to solve the multicriterion optimization (MCE) problem. A relatively new and progressive class of MCE methods are methods that presuppose the preliminary construction of some finite-dimensional approximation of a set of solutions called Pareto-approximation. At present, there are quite a few Pareto-approximation methods, which raises the question of a comparative evaluation of the effectiveness of these methods. As shown in the previous works of the authors, many software systems implementing these algorithms do not provide fast access and do not have an easy-to-use interface for calculations and analysis. In this paper, the development of a service-oriented system for the performance and evaluation of the Pareto-approximation quality is considered. The main purpose of this system is to provide fast and convenient access to Pareto-approximation solution and evaluation algorithms, without attachment to a particular operating system or application.

When solving practical, primarily, technical optimization tasks, the tasks of multi-objective optimization (MCHO-tasks) often arise, when it is desirable to exaggerate not one but several objective functions. It is well known that since, as a rule, it is impossible to achieve the extreme value of all objective functions simultaneously, the solution of such a problem is a certain compromise belonging to the Pareto set of the problem, that is, to a set of solutions that are not simultaneously improved for all objective functions. By the solution of the MCE-problem, we call the Pareto set of this problem, assuming that, after constructing some finite-dimensional approximation of the given set (P-approximation), the person making the decision (LPR), using informal or formalized methods, selects the unique solution from it. WEB-oriented software system PARETO-Q is designed for remote evaluation of Pareto-approximation quality in multi-purpose optimization problems. The system is designed to solve the problem of consolidation of Pareto-approximation algorithms and algorithms for estimating the quality of approximation. We represent the software used by the system, the architecture of the system, some results of computational experiments.

pp. 150-164

pp. 197-214

pp. 193-214

The paper makes a statement of the multi-criteria optimization problem, gives a brief review of techniques to solve it using its Pareto-approximation, and presents mainly known approximation quality indicators. Principally, the paper gives a review of known programme systems orientated to the Pareto-approximation problem solution, which implement various algorithms to estimate the quality of this approximation.