National Technical University of Athens (NTUA)
Υβριδικές ευφυείς τεχνικές ανάλυσης για εξόρυξη γνώσης από δεδομένα υψηλών διαστάσεων
Abstract
dc:descriptionThe research on hybrid systems of artificial intelligence and machine learning models presents an increasing interest during recent years. The contribution of the current thesis is focused on this field by developing novel models and learning methods and designing original hybrid systems. More specific, initially, two different clustering methods have been implemented, based on the common idea of using first a simple clustering algorithm for the discovery of an initial group of clusters. In the first method, the FCM algorithm is used while in the second the self-organizing maps model. Following, in both cases, an agglomerative clustering algorithm is applied in order to achieve a more efficient clustering, each time with the necessary modifications so as to take advantage of meta-data deriving from the initial clustering procedure. With the use of the second clustering procedure, a feature evaluation method has been developed. This method is the first part of a hybrid classification system and the results of the method are fed to a neuro-fuzzy classifier, which is the next part of the system. Another hybrid system has also been developed. The main module of this system is a novel memory-based classifier which interacts with a feature evaluation module that is based on the self-organizing maps model. The self-organizing maps model has been also employed for implementing transformations of extracted knowledge to symbolic form. The first approach of this kind has been the development of a hand gesture recognition system that uses self-organizing maps for generating symbolic states so as to create probabilistic classification models. The second approach is a group of rule extraction methods from a trained self-organizing map. The performance of all the above methods and systems has been tested on benchmark datasets, which are widely used by researchers in the corresponding-fields.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2007
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Pateritsas, Christos
- Πατερίτσας, Χρήστος
Subjects
dc:subject × 16- Υβριδικές τεχνικές
- Ευφυή συστήματα
- Δίκτυα, Νευρωνικά
- Εξόρυξη γνώσης
- Μηχανική μάθηση
- Ανάλυση δεδομένων
- Hybrid method
- Intelligent systems
- Networks, Neural
- Knowledge extraction
- Machine learning
- Data analysis
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/17041