Athens University Economics and Business (AUEB)
Real-time pattern recognition in data streams in the internet of things
Abstract
dc:descriptionThe rapid growth of the Internet of Things, coupled with blockchain-based and traditional financial transactions and social networks, has led to the generation of vast data streams, which include an unprecedented amount of information that are destined to be lost to the oblivion unless they are analyzed and processed optimally, without the hard requirement for them to be stored. This dissertation addresses the challenges and opportunities associated with analyzing such data streams, emphasizing the need for methodologies that can handle continuous data influx without relying on vast amounts of stored historical data. Traditional machine learning algorithms often fall short in this context due to their need for multiple data passes over the whole dataset, inability to adapt to changes in the statistical properties of the data streams and high computational demands. This work explores the adaptation of various traditional clustering and classification algorithms for real-time data stream analysis, through the patented BEReTiC methodology. BEReTiC is a real-time clustering and classification algorithm multiplexer and adapter, designed to adapt to the unique properties of data streams, such as concept drift and evolution. The algorithm CluNN, a novel patented clustering algorithm, that has been specifically designed for data stream clustering is also presented. The performance and the accuracy of the methodology and the algorithm is assessed through qualitative and empirical evaluations across multiple real-world domains, including applications characterized by data streams with unknown or volatile distributions and applications whose data streams abide by more predictable distributions. The results demonstrate that BEReTiC and CluNN enable timely, adaptive and accurate pattern recognition in both volatile and deterministic environments, addressing critical challenges in real-time data stream analysis. These contributions pave the way for modular, efficient, and practical integration of pattern recognition capabilities in real-world IoT systems.
Degree
thesis:*- Grantor dc:publisher
- Athens University Economics and Business (AUEB)
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Polytarchos, Elias
- Πολυτάρχος, Ηλίας
Subjects
dc:subject × 27- Μηχανική μάθηση
- Αλγόριθμοι ομαδοποίησης και ταξινόμησης
- ΜΜη επιβλεπόμενη στατιστική ανάλυση δεδομένων
- Ροές δεδομένων
- Αναγνώριση προτύπων
- Διαδίκτυο των πραγμάτων
- Ανάλυση δεδομένων σε πραγματικό χρόνο
- Εξόρυξη δεδομένων σε πραγματικό χρόνο
- Μεθοδολογίες προσαρμογής αλγορίθμων
- Ομαδοποίηση σε πραγματικό χρόνο
- Machine learning
- Clustering and classification algorithms
- Unsupervised statistical data analysis
- Data streams
- Pattern recognition
- IoT (Internet of Things)
- Real-time data analysis
- Real-time data mining
- Algorithm adaptation methodologies
- Clustering and classification algorithm ensembles
- Real-time clustering
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Τεχνητή νοημοσύνη
- Computer and Information Sciences
- Natural Sciences
- Artificial Intelligence
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59764