Ionian University
Αυτοματοποιημένη κατηγοριοποίηση δυναμικών δεδομένων με έμφαση στις σελίδες διαδικτύου: μια συνδυαστική προσέγγιση
Abstract
dc:descriptionThis dissertation deals with the automated dynamic data classification, and in particular with the web pages’ automated classification through a combined approach. Since the internet has appeared, this field is widely studied given that key aspects of computer science rely on web pages’ classification, such as information management and retrieval, information interoperability, and also information extraction models, content filtering models and so on. Recently, the significant development of the computer performance and memory space, combined with the machine learning specialization models for text and image classification, are further reasons why the web pages’ classification remains at the center of research interest. At the same time, additional challenges are the complexity of automated web pages’ classification as a process, the diversity of web page content (images of different sizes, text, hyperlinks, etc.) and the cost of computing. Looking up the World Wide Web content management approaches presented in the international literature, we find that most of them rely mainly on text categorization techniques and some others utilize the pages’ structure. This dissertation aims to design a hybrid approach for the problem of automated web pages’ classification, based on both textual elements and structural features. In other words, the proposed approach is built on existing works, combining techniques used in their context so that the pages are categorized in terms of their topic, but also in terms of their type. This means that the proposed approach is a combined unified web page classification methodology, which exploits web pages’ textual and structural elements, features and characteristics.To this end, the proposed methodology assigns every page examined with the appropriate category, after having checked various parameters related to the pages’ content and structure. From this point of view, we design a multidimensional categorization algorithm, which decides on the type and topic of every examined page. Additionally, looking closer the web pages and their dynamic nature, we extend the potential of the proposed methodology, including two additional algorithms, which monitor, detect and control the need to re-classify the web pages, where necessary, based on changes that may have occurred in content and/or structure. In this way, it is introduced a regular check of the classified pages, so that the classification result is always updated.Concerning the effectiveness and the efficiency of our methodology, we carry out a pilot experimental evaluation. Through this process it is shown that web pages are correctly classified in two ways; according to the theme of their content and according to their structural type. To boost the evaluation of the proposed methodology, we additionally carry out a comparative study between the results of the proposed classification algorithm and the results obtained by the application of a k-NN algorithm. This comparative study shows that the performance of the proposed algorithm can be compared and is equivalent to the performance of a traditional text classification algorithm.
Degree
thesis:*- Grantor dc:publisher
- Ionian University
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Niarou, Maria
- Νιάρου, Μαρία
Subjects
dc:subject × 20- Διαχείριση δεδομένων
- Κατηγοριοποίηση
- Δυναμικά δεδομένα
- Δομικά χαρακτηριστικά
- Σημασιολογικά δεδομένα
- Οντολογίες
- Αλγόριθμοι
- Data management
- Classification
- Dynamic Data
- Structure properties
- Semantic data
- Ontologies
- Algorithms
- ΜΜΕ και Επικοινωνίες
- Κοινωνικές Επιστήμες
- Επιστήμη πληροφόρησης
- Media and Communications
- Social Sciences
- Information science
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/50925