Aristotle University Of Thessaloniki (AUTH)
Εκτίμηση δικτύου συνδεσιμότητας πολύπλοκων συστημάτων από χρονοσειρές υψηλής διάστασης
Abstract
dc:descriptionThis PhD thesis deals with connectivity network estimation from high-dimensional time series with applications in the field of finance. In the first and introductory chapter, reference is made to the ever-increasing presence of time series in the modern era and their analysis methods. In the second chapter, basic concepts related to dynamic, deterministic and stochastic systems are mentioned, while in the third chapter, the way of applying network theory to the analysis of multivariate time series is briefly presented.In the fourth chapter, the estimation of causality networks from high-dimensional time series is investigated using non-linear information measures, which are based on Granger causality estimation, while the influence of network density on the behavior of these measures is also studied. In the fifth chapter, a methodology for detecting the emergence or disappearance of non-linear interactions in complex systems is presented based on an appropriate statistical hypothesis test. An algorithm for detecting structural changes in complex systems is also developed, which is based on the application of the aforementioned methodology in time, rolling windows. In the sixth chapter and in continuation of the previous results, improvements are proposed for the non-linear Granger causality estimation measure of partial mutual information from mixed embedding which aim to improve weaknesses of the measure that have been observed in previous studies and mainly concern cases of strongly stochastic systems.In the seventh chapter, canonical correlation analysis (CCA), Louvain's algorithm from network theory and a randomization test are used to identify communities in correlation matrices from high-dimensional multivariate time series. The eighth and last chapter concludes the thesis by gathering its main conclusions.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Fotiadis, Akilas
- Φωτιάδης, Ακύλας
Subjects
dc:subject × 22- Υψηλοδιάστατες χρονοσειρές
- Δίκτυα αιτιότητας
- Μέτρα πληροφορίας
- Μείωση διάστασης
- Ανίχνευση δομικών αλλαγών
- High-dimensional time series
- Causality networks
- Information-based measures
- Dimension reduction
- Structural breaks detection
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Τεχνητή νοημοσύνη
- Οικονομικά και Επιχειρήσεις
- Κοινωνικές Επιστήμες
- Χρηματοοικονομικά
- Computer and Information Sciences
- Natural Sciences
- Artificial Intelligence
- Economics and Business
- Social Sciences
- Finance
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/53683