Повышение качества идентификации случайных временных рядов на основе модифицированных схем мяхких вычислений
- № 3(15) 2018
Страницы:
96
–
110
Язык: русский
Аннотация
Сформулирована проблема, предложены подходы, решены задачи повышения достоверности прогноза случайных временных рядов (СВР) при неизвестных, недостаточно заданных априорных параметрах в условиях влияния внешней среды на каналыпереработкиинформации.Разработаныметодыоптимизацииобработкиданных, использующие аппарат нечеткого моделирования, автоматического регулирования и настройки значений переменных. Решены задачи построения технологии идентификации, оптимизации, обработки данных. Определены характерные исследуемым процессам функции принадлежностей (ФП), их параметры, границы, интервалы принадлежности носителя соответствующим нечетким множествам лингвистических терм входных и выходных переменных. Разработаны механизмы регулирования и настройки значений элементов СВР, определения и реализации стратегии оценки достоверности обработки данных и принятия альтернативных решений. Разработана обобщенная вычислительная схема нечеткой идентификации СВР, механизмов регулирования и настройки элементов, реализованных по принципам прямой и инверсной связей. Для оптимизации реализованы принципы нечеткой идентификации объектов. Реализованный программный комплекс протестирован для прогнозирования технико-экономических показателей производственных предприятий.
The problem is formulated, approaches are proposed, tasks are solved for increasing the reliability of forecast of random time series (RTS) with unknowns, not enough given priori parameters under the influence of environment on information processing channels. Methods are developed to optimize data processing using apparatus of fuzzy modeling, automatic control and adjustment variables. The tasks are solved for constructing the technology of identification, optimization and data processing, as well as computational schemes for establishing statistical and specific characteristics, dynamic properties of information, regularities, useful knowledge in data, features of objects. The belonging functions (BF), characteristic to investigated processes, their parameters, borders, spacing, intervals of carriers accessories to corresponding fuzzy sets of linguistic terms of input and output variables. The efficiency of mechanisms to forming the fuzzy terms sets, establishing the optimum boundaries and BF parameters for each segment, the carrier interval of according fuzzy sets of linguistic terms. Mechanisms for regulating and adjusting the values of RTS elements, defining and implementing a strategy for evaluating the reliability of data processing, and adopting alternative solutions are developed. A generalized computational scheme for fuzzy identification of RTS, mechanisms for regulating and adjusting of elements implemented according to the principles of direct and inverse relationships is developed. Foroptimizingtheprincipleoffuzzyidentificationofobjectarerealizedonthebasisof operators and procedures for logical inclusion, equivalence, union, intersection, addition, algebraic sum calculations, product, exponentiation, center concentration, defining of coefficients and borders of BF, stretching and contraction of BF boundaries, choice of BF type, intervals of belonging of fuzzy sets carriers. The realized software complex is tested for forecasting technical and economic indicators of product corporations.
The problem is formulated, approaches are proposed, tasks are solved for increasing the reliability of forecast of random time series (RTS) with unknowns, not enough given priori parameters under the influence of environment on information processing channels. Methods are developed to optimize data processing using apparatus of fuzzy modeling, automatic control and adjustment variables. The tasks are solved for constructing the technology of identification, optimization and data processing, as well as computational schemes for establishing statistical and specific characteristics, dynamic properties of information, regularities, useful knowledge in data, features of objects. The belonging functions (BF), characteristic to investigated processes, their parameters, borders, spacing, intervals of carriers accessories to corresponding fuzzy sets of linguistic terms of input and output variables. The efficiency of mechanisms to forming the fuzzy terms sets, establishing the optimum boundaries and BF parameters for each segment, the carrier interval of according fuzzy sets of linguistic terms. Mechanisms for regulating and adjusting the values of RTS elements, defining and implementing a strategy for evaluating the reliability of data processing, and adopting alternative solutions are developed. A generalized computational scheme for fuzzy identification of RTS, mechanisms for regulating and adjusting of elements implemented according to the principles of direct and inverse relationships is developed. Foroptimizingtheprincipleoffuzzyidentificationofobjectarerealizedonthebasisof operators and procedures for logical inclusion, equivalence, union, intersection, addition, algebraic sum calculations, product, exponentiation, center concentration, defining of coefficients and borders of BF, stretching and contraction of BF boundaries, choice of BF type, intervals of belonging of fuzzy sets carriers. The realized software complex is tested for forecasting technical and economic indicators of product corporations.