The basis patterns method for non-stationary time-series analysis
Abstract:
The definitions of approximate expansion of empirical distribution function density over special basis patterns are proposed for non-stationary time series. The method of statistical recognition of the local state in terms of distribution function is formulated. The statistics of trend prediction is constructed.
Keywords:
non-stationary time series, basis patterns, self-consistent stationary level, pattern recognition
Publication language:russian, pages:20
Research direction:
Mathematical modelling in actual problems of science and technics
