|
Title:
|
MONITORING FRACTAL PROPERTIES OF TIME SERIES VIA HURST EXPONENT ESTIMATION AND LSTM-BASED DETECTION |
|
Author(s):
|
Lyudmyla Kirichenko, Sergiy Yakovlev, Alexander Kirpich, Olha Matsyi and Dmytro Chumachenko |
|
ISBN:
|
978-989-8704-71-9 |
|
Editors:
|
Paula Miranda and Pedro Isaías |
|
Year:
|
2025 |
|
Edition:
|
Single |
|
Keywords:
|
Hurst Exponent, Fractal Analysis, LSTM Autoencoder, EEG, Anomalous Diffusion, Time Series |
|
Type:
|
Poster |
|
First Page:
|
318 |
|
Last Page:
|
320 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
if you are a member please login 
|
|
Paper Abstract:
|
Many technological, physical, and biological systems exhibit complex fractal behavior. Detecting changes in the fractal structure of time series is essential, as they may indicate transitions to anomalous or critical states. This study proposes a method for tracking such changes by monitoring the Hurst exponent using a sliding window and the Whittle estimator. To detect significant variations in the Hurst parameter, we apply LSTM autoencoders, which can capture temporal dependencies and identify deviations via reconstruction error. The model was trained on synthetic fractal time series with various types of Hurst dynamics, including abrupt shifts and gradual trends. Validation on both synthetic and real EEG data demonstrated the method's ability to detect meaningful structural changes. |
|
|
|
|
|
|
