University of Piraeus (UNIPI)
Design, optimization, development and explanation of artificial intelligence algorithms, capable of running on distributed big data systems, based on multidimensional and complex datasets
Abstract
dc:descriptionThe exponential growth of digital data and advancements in Artificial Intelligence (AI) have redefined data-driven decision-making across industries. However, managing complex, high-dimensional, and multi-frequency datasets remains challenging, particularly in finance and cloud computing sectors where data variability, real-time adaptability, and predictive precision are crucial. This thesis introduces data-centric AI methodologies designed to create resilient and adaptable systems capable of processing diverse data across multiple formats and domains. Specifically, the dissertation presents a Knowledge and Reasoning Framework that utilizes ontologies and retrieval-augmented generation (RAG) techniques for integrating, managing, and processing structured and unstructured data. This enables the combined utilization of different types of data originating from different sources. Furthermore, techniques are developed to improve the real-time adaptability of time-series analysis models, allowing continuous system updates to capture evolving trends and maintain robust performance in dynamic environments. Techniques are also presented for the optimized use of Large Language Models (LLMs) for text analysis and classification, comparing their effectiveness with traditional transfer learning techniques for sentiment analysis in financial articles. Finally, an innovative system is proposed that combines and extends these methodologies, managing text, numerical data, and time-series data from various sources, providing a comprehensive solution with applications in the financial sector. The dissertation validates the proposed methodologies through real-world applications such as time-series prediction for cloud computing resource management, financial risk assessment, sentiment analysis, and systematic stock selection in finance. These contributions demonstrate significant improvements in handling complex datasets, ensuring real-time adaptability, and delivering interpretable and reliable AI-driven insights. By addressing the common challenges of data complexity and heterogeneity in real-world applications through multiple data-centric methodologies, this thesis offers scalable and resilient AI solutions for data-rich environments, effectively managing diverse and complex data formats across various domains.
Degree
thesis:*- Grantor dc:publisher
- University of Piraeus (UNIPI)
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Fatouros, Georgios
- Φατούρος, Γεώργιος
Subjects
dc:subject × 28- Τεχνητή νοημοσύνη
- Μεγάλα γλωσσικά μοντέλα (LLMs)
- Χρηματοοικονομική ανάλυση
- Εξόρυξη γνώσης
- Πρόβλεψη χρονοσειρών
- Υπολογιστικό νέφος
- Μηχανική μάθηση
- Βαθιά μάθηση
- Ανάκτηση βάσει περιεχομένου
- Artificial intelligence
- Large language model (LLMs)
- Financial forecasting
- Data mining
- Time series
- Cloud computing
- Machine learning
- Deep learning
- Retrieval of context
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Ρομποτική
- Computer and Information Sciences
- Natural Sciences
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Robotics
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/58200