Nonparametric analysis of stochastic systems with nonlinear functional heterogeneity

The problems of the analysis of stochastic systems described by nonlinear statistical models with heterogeneous functional forms are considered in the space of «essential dependent» features. It is supposed that functional heterogeneity is conditioned by the existing of the different classes of system states. The algorithm of classification of the systems states as well as the forecasting algorithm for endogenous variables based on multivariate nonparametric density estimate with adaptive kernel are described and examined by means of statistical modeling experiments