Time series classification based on fractal properties

TL;DR

This paper explores classifying fractal time series using decision tree-based meta algorithms, demonstrating machine learning methods outperform traditional self-similarity measures in accuracy.

Contribution

It introduces a machine learning approach for fractal time series classification and compares it with traditional methods, highlighting its advantages.

Findings

Machine learning methods outperform traditional self-similarity estimation.

Decision tree-based meta algorithms effectively classify fractal time series.

Comparative analysis shows improved accuracy with ML approaches.

Abstract

The article considers classification task of fractal time series by the meta algorithms based on decision trees. Binomial multiplicative stochastic cascades are used as input time series. Comparative analysis of the classification approaches based on different features is carried out. The results indicate the advantage of the machine learning methods over the traditional estimating the degree of self-similarity.