Wavelet-Based Estimation of Hurst Exponent Using Neural Network
Lyudmyla Kirichenko, Kyrylo Pavlenko, Daryna Khatsko · 2022 IEEE 17th International Conference on Computer Sciences and Information Technologies (CSIT) · 2022
The paper proposes a method for estimating the Hurst exponent of time realizations using a regression neural network. The basis was the wavelet estimation method using a discrete wavelet transform. Wavelet energy spectra of fractional Brownian motion realizations were fed to the input of the neural network. The results showed that the accuracy of the Hurst exponent estimation, performed using a neural network, is ten times higher than the accuracy of statistical wavelet-based estimation.