A study on rough set-aided feature selection for automatic web-page classification

Article type: Research Article

Authors: Wakaki, Toshiko | Itakura, Hiroyuki | Tamura, Masaki | Motoda, Hiroshi | Washio, Takashi

Affiliations: Shibaura Institute of Technology, 307 Fukasaku, Minuma-ku, Saitama-City 337-8570, Japan | Japan Advanced Institute of Science and Technology, 1-1, Asahidai, Tatsunokuchi, Ishikawa 923-1292, Japan | The Institute of Scientific and Industrial Reasearch, Osaka University, Mihogaoka, Ibaraki, Osaka 567-0047, Japan

Note: [] Corresponding author. Tel.: +81 48 687 5158; Fax: +81 48 687 5198; E-mail: [email protected]

Abstract: Due to the recent explosive increase of Web-pages on World Wide Web, it is now urgently required for portal sites like Yahoo! service having directory-style search engines to classify Web-pages into many categories automatically. This paper investigates how rough set theory can help select relevant features for Web-page classification. Our experimental results show that the combination of the rough set-aided feature selection method and the Support Vector Machine with the linear kernel is quite useful in practice to classify Web-pages into multiple categories because not only our experiments give acceptable accuracy but also the high dimensionality reduction is achieved without the need to search for a threshold for feature selection.

Keywords: Rough set, feature selection, Web-page classification, Support Vector Machines, C4.5, dimensionality reduction

Journal: Web Intelligence and Agent Systems: An international journal, vol. 4, no. 4, pp. 431-441, 2006