About algorithm of robust nonparametric estimation of regression function on observations

9th International Multidisciplinary Scientific and Research Conference "Modern Issues in Science and Technology" Workshop "Advanced Technologies in Aerospace, Mechanical and Automation Engineering"

自然科学;工业技术

Denisov, M.A.^1 ; Chzhan, E.A.^1 ; Korneeva, A.A.^1 ; Kukartsev, V.V.^1,2

Siberian Federal University, Institute of Space and Information Technologies, Krasnoyarsk

660074, Russia^1

Siberian State Aerospace University, Krasnoyarskiy Rabochiy Ave., 31, Krasnoyarsk

660037, Russia^2

关键词: Identification method;    Inter quartile ranges;    Non-parametric estimations;    Non-parametric identification;    Parametric method;    Priori information;    Regression function;    Robust statistics;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/450/4/042001/pdf
DOI  :  10.1088/1757-899X/450/4/042001

来源: IOP

【 摘 要 】

The field of research presented in the paper is aimed at studying the methods of robust statistics for the modeling of multidimensional processes of discrete-continuous type. The model of the investigated process is constructed using identification methods. In one case, it is used parametric methods of identification where priori information about the object of research is sufficient to build the model accurate within vector of parameters. The second case is specific to lack of priori information. The researchers do not know the structure of the model and represent the object in the form of a "black box", therefore, nonparametric identification methods could be used. The accuracy of models is estimated using a relative error of approximation, which shows how much the model output value corresponds to the output value of the object. The paper proposes a new method of outliers' detection in the initial sample of observations, which is subsequently used for parametric and nonparametric identification of processes. The developed robust algorithm is applied to both types of models in order to determine in which case accuracy of outliers' detection is higher. In addition, the above algorithm is compared with an algorithm based on the interquartile range.