Оцінка показника Херста на основі вейвлетів за допомогою нейронної мережі

Abstract

This work is devoted to an approach to estimating the Hurst exponent for fractal time realizations, based on the application of machine learning, namely a regression neural network. The construction of the regression model is based on the estimation of the Hurst index using a discrete wavelet transformation. The input data are the values of the wavelet energies obtained from realizations of the fractal Brownian motion. Comparative analysis showed that when using a neural network, the accuracy of the assessment increased several times.