Исследуется задача восстановления матрицы наблюдений при оценивании функции регрессии по измерениям со случайными ошибками. Заполнение пропусков осуществляется с помощью непараметрической оценки кривой регрессии. Приводятся результаты численных исследований, иллюстрирующих эффективность работы предложенной методики. Рассматривается моделирование нового класса процессов стохастических объектов со статистической зависимостью компонент вектора входа.
By investigation of many processes it is necessary to solve the modeling and identification problem. The qualitatively constructed models help to simplify the control of the object as well as to predict its future behavior. This paper focuses on the identification of a new class of processes which have statistical relationship between the components of the input variables. Further, these objects will be called «tubular». As it is known, the quality of the identification problem solution is determined by the quality of source data, so the stage of data preprocessing is an important part of the modeling process. In this paper some peculiarities of the samples, as blanks are described. There are proposed two nonparametric estimation algorithms using the regression function. The use of parametric identification methods does not give satisfactory results in modeling «tubular» processes. A modification of the parametric identification algorithm using the indicator function is suggested. The indicator shows whether the points belong to the true course of the process or not. The experimental results demonstrate the feasibility of the proposed algorithms.