Komplexní sítě pro analýzu časových řad

Abstract

This bachelor thesis explores the possibilities of utilizing complex networks for the purpose of time series analysis. For this time series processing, selected methods of complex networks are implemented in the Python programming language. For testing complex networks, a time series of natural gas consumption for the time periods 2015 and 2019 is chosen with aggregation on a daily and monthly basis. This processing is then visualized using several types of graphs, where communities, centrality, anomalies, and important nodes are shown. It turned out that the visibility graph appears to be the most advantageous for representing the time series of gas consumption. Among other options, the electrostatic graph and the transition graph can be used, where the representation of the complex network depends on the correct configuration. On the other hand, the permutation graph and the quantile graph proved to be unsuitable.