Models and procedures for classification and forecasting of nondeterministic processes according to chaotic dynamics parameters

Abstract

The article investigates the processes of classification, modeling and short-term prediction of nondeterministic time sequences, which are represented by antipersistent time series (ATS). The subject of analysis - procedures for classification and forecasting the pa-rameters of such models. The object of research is the processes of modeling and analysis of parameters of nondeterministic time series of ATS with a uniform step. The aim of the work is to increase the efficiency and accuracy of methods and algorithms for classification, modeling and forecasting of ATS. Models and methods of fractal analysis are used to study the properties of ATS, on the basis of which the categories of processes of numerical series are established. With the help of aggregation of ATS levels correct mathematical models of classification of nondeterministic time sequences are developed, and also algorithmic and software means of their realization are formed. Examples of models of numerical series obtained using the aggregation procedure presented in the study are given. It is established that the most detailed and stable is the classification of ATS based on data aggregation schemes without level crossing. The comparative analysis of numerical efficiency of algorithms of classification of ATS is carried out and the task of formation of procedures of interpolation and short-term forecasting of ATS is realized. An instrumental software environment is presented, which provides a correct study of algorithms for modeling and classification of antipersistent time series. Recommendations on the procedures for modeling ATS classification algorithms are of practical importance.