The Analysis Of Metadata Based Classification For Classifying Educational Websites

• Main Author: Zaraini, Mohd Nazrien
• Format: Thesis
• Language: English; English
• Published: 2016
• Subjects: Z665 Library Science. Information Science
• Online Access: http://eprints.utem.edu.my/id/eprint/18198/; https://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=100103

Initially websites can be easily categorized based on its domain extensions. But due to the explosion of the internet, the domain name restrictions are no longer being adhered. Web classification can help to categorize websites, especially educational websites that being the focus of this research. Classification will be done based on content and metadata in order to get the impact of metadata implementation in terms of classification accuracy. Three sets of 200 pre-determined educational websites taken from DMOZ directory utilized as training data. This is the total number of educational websites with metadata information available in that directory. For content based classification, keywords extracted from the contents and TF-IDF ranking used to get the top educational keywords. These keywords used as a training dataset attribute for educational web classification. The same method goes for metadata based classification, but the difference is that the keywords were taken from its meta description. One class support vector machine method was used because this research is focusing on single class classification only. Cross validation technique and two sets of test data; all educational websites and various categories of website will be used to validate this research. The results shows that content based classification gives more accuracy compare to metadata. Top ranking educational keywords and the analysis of metadata implementation known from this research based on the information retrieval and web classification process.