The non-parametric algorithm of omissions filling in stochastic data

Abstract

The paper presents the results of an algorithm for data processing. In the initial data omissions may occur, due to different control discreteness of input and output variables. The paper proposes a non-parametric algorithm for filling gaps. The basic idea is to calculate the non-parametric estimate of the regression function from observations obtained from the object. This allows to use all available measurements. Numerous computational experiments have shown that the use of the proposed algorithm has improved the quality of the resulting model several times. The algorithm is influenced by such parameters as the total number of omissions in the sample of observations, measurement interference in communication channels, and the type of object. It should be noted that the developed algorithm is universal and does not depend on the type of equation of the object.