Nonparametric estimation of regression curves in the conditions of large samples

The technique of construction of a nonparametric regression in the conditions of training samples of large volume is offered. Model synthesis is based on decomposition of initial statistical data and the analysis of probabilistic characteristics of received random variables sets. Asymptotic properties of a nonparametric regression are investigated and results of computing experiment are considered. Association of nonparametric regression properties on an amount of sampling intervals of values of an random variable and volume of input datas is established. Comparison of approximating properties of offered model and a traditional nonparametric regression is spent. The results of researches are important to the solution of problems of a confidential estimation of a probability density and a regression curves.