ИДЕНТИФИКАЦИЯ ПАРАМЕТРОВ КУСОЧНО-ЛИНЕЙНОЙ ФУНКЦИИ РИСКА МЕТОДОМ АНТИРОБАСТНОГО ОЦЕНИВАНИЯ

Abstract

The article reviews literature on identification methods for parameters of regression models, including those based on Chebyshev metrics. The publications contain data on the development of an algorithm for the unambiguous definition of Chebyshev projection; a new method that combines Minkowskian distance and Chebyshev distance, with both being used as a similarity measure in the clustering process when grouping data; generalizations of particular goal setting for curves and surface fitting to the data observed or calculated as a result of replacing least squares with the Chebyshev norm; and integral estimates of the anthropogenic transformation of the territory using multidimensional statistical methods. Using the anti-robust method of estimation, the authors have developed a method to estimate unknown parameters of a regression piecewise linear risk function, whose aim is to solve a linear Boolean programming problem. The risk function of the dynamics of the price per square meter of living space of dwellings in the housing market in the Russian Federation is built using the least modules method and the anti-robust estimation method. Average pricing for silicate wall blocks, concrete floor slabs, and ready-mixed concrete are used as independent factors in the model. Both versions of the models built describe the dynamics of the output indicator profoundly, as evidenced by the high values of adequacy criteria, and therefore can efficiently solve the forecasting problems. It has been established that the number of maximum module errors of the risk model approximation is equal to three, i. e. the number of independent variables, when applying the anti-robust method.