РАЗРАБОТКА ГИБРИДНОЙ СИСТЕМЫ ПОДДЕРЖКИ ПРИНЯТИЯ РЕШЕНИЙ И ЕЕ ПРИМЕНЕНИЕ
In this paper author suggests a new hybrid decision support system for operation with a class of semistructured tasks with underdetermined variables. Author defined the general tasks of prediction and estimation for a class of semistructured tasks. Use of interval neural networks and genetic algorithms for such tasks is justified. Author developed the algorithm to train interval neural networks. The diagram of the offered decision support system is described. Use of technologies for parallel computation on GPU kernels is justified. Author developed an effective algorithm of the developed algorithms parallel computation. Two examples of use of the developed system are given: prediction of the sea ice area in the Northern hemisphere and prediction of client solvency for credit institutions.