EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

GRAPH BASED DETECTION OF FUTURE PATTERNS IN TIME SERIES

Maria Katsouda, Basilis Boutsinas

Far East Journal of Mathematical Sciences (FJMS) · 17.07.2026 · 143 (9) · 2965-2991

DOI: 10.17654/0972087126164 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Far East Journal of Mathematical Sciences (FJMS), online erschienen am 17.07.2026. Verfasst von Maria Katsouda und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Time series often exhibit recurring patterns, the identification of which can provide important information about the behavior of the systems that generate them, particularly in data-driven business environments where access to reliable information is essential for strategic decision making. Many such approaches have been proposed in the literature, mainly for forecasting, which are based on analyzing almost the entire time series and are also dependent on parameters that must be set manually. In this study, we propose a pattern-matching algorithm that automatically converts a univariate time series into a graph based symbolic representation capable of predicting sequential patterns. The graph can be built based on only a small part of a time series while the remaining part of the series is used to verify it. Thus, the proposed algorithm can support a short-term prediction on the time series. To build the graph, consecutive numerical differences between values, within a sliding window of the time series of length n, are likewise transformed into symbolic ones and inserted into the graph. In the built directed graph, nodes symbolically represent values of the time series and edges represent differences between consecutive such values. The graph is used to evaluate the validity of future subsequences: a future sequence is considered valid if its corresponding symbolic pattern already exists in the graph, and not valid otherwise. The proposed approach contributes to Financial Mathematics and to Business Intelligence problems by enabling automated pattern detection and short-term forecasting in financial time series. Experimental results on real-world and financial data confirm its effectiveness for decision support.

Zitieren

APA

Katsouda, M., & Boutsinas, B. (2026). GRAPH BASED DETECTION OF FUTURE PATTERNS IN TIME SERIES. Far East Journal of Mathematical Sciences (FJMS), 143(9), 2965-2991. https://doi.org/10.17654/0972087126164

Vancouver

Katsouda M, Boutsinas B. GRAPH BASED DETECTION OF FUTURE PATTERNS IN TIME SERIES. Far East Journal of Mathematical Sciences (FJMS). 2026;143(9):2965-2991. doi:10.17654/0972087126164

BibTeX
@article{Katsouda2026,
  author = {Katsouda, Maria and Boutsinas, Basilis},
  title = {GRAPH BASED DETECTION OF FUTURE PATTERNS IN TIME SERIES},
  journal = {Far East Journal of Mathematical Sciences (FJMS)},
  year = {2026},
  volume = {143},
  number = {9},
  pages = {2965--2991},
  doi = {10.17654/0972087126164}
}
Kostenlos starten →