Aristotle University Of Thessaloniki (AUTH)
Στατιστικές μέθοδοι ταξινόμησης πολυμεταβλητών δεδομένων
Abstract
dc:descriptionThe present dissertation deals with the investigation and tackling of several issues that arise frequently in classification and prediction of multivariate data. Specifically, these issues are related to the selection of features / variables and cases, to the well-known class imbalance problem and generally, to factors that affect the classification procedure. The goal of the dissertation is to find and present suitable approaches and methods that can improve the performance of classification methods. All the aforementioned issues are interesting research topics of the Mahalanobis-Taguchi (MT) strategy, a rather new methodology that incorporates various mathematical and statistical concepts, such as Mahalanobis distance (MD), Gram-Schmidt orthogonalization process and experimental designs. The special characteristic of the MT strategy is that it is not only a classification technique, but also a toolbox of methods, the different components of which have not been studied thoroughly in the relevant literature. Hence, the contribution of the dissertation mainly focuses on investigating and resolving the operational issues of the MT strategy. The dissertation consists of two sections: a) the improvement of the general MT strategy’s procedures by employing other statistical methods and b) the proper adaptation of the methodology, in order to be applicable to different types of data.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2015
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Liparas, Dimitrios
- Λιπαράς, Δημήτριος
Subjects
dc:subject × 12- Στατιστική ταξινόμηση
- Στρατηγική Mahalanobis-Taguchi
- Πρόβλεψη ατελειών λογισμικού
- Ανάλυση εγκεφαλογραφικών αποκρίσεων
- Statistical classification
- Mahalanobis-Taguchi strategy
- Software defect prediction
- Analysis of encephalographic responses
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Natural Sciences
- Computer and Information Sciences
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/35839