Increasing The Credibility Of Forecasting Random Time Series Based On Fuzzy Inference Algorithms

Abstract

Methods for the identification of linear, nonlinear dependencies help models of fuzzy logic and neural network (NN), data preprocessing, design of a computational training scheme for a five-layer neuro fuzzy network (NFN) are proposed. A software and algorithmic complex has been implemented, including modules for computational circuits of the NFN, parametric and structural identification. The effectiveness of methods for forecasting random time series is shown using the example of numerical results.