KATEGORISASI UNBALANCED TEXT MENGGUNAKAN COMPLETE GINI-INDEX DAN RELATIVE WEIGHT K-NEAREST NEIGHBOR

Widiasri, Monica and Justitia, Army (2013) KATEGORISASI UNBALANCED TEXT MENGGUNAKAN COMPLETE GINI-INDEX DAN RELATIVE WEIGHT K-NEAREST NEIGHBOR. In: Seminar Nasional Teknologi Informasi dan Multimedia (SNASTIA) 2013, Universitas Surabaya ISSN 1979-3960, 21 September 2013, Universitas Surabaya.

[img]
Preview
PDF
SNASTIA_Abstrak_2013_MonicaW_Kategorisasi.pdf

Download (581Kb) | Preview
[img] PDF
SNASTIA_2013_MonicaW_Kategorisasi.pdf
Restricted to Registered users only

Download (1206Kb)

Abstract

Feature selection is necessary to reduce a large feature space in text categorization. Especially in unbalanced text, where the number of training documents between each category is unbalanced,it needs proper feature selection method that features selected are appropriate and can distinguish between categories. Also, a proper categorization method is needed to categorize unbalanced text, because unsuitable categorization method for unbalanced text categorization can lead poor results. This research used Complete Gini-Index (CGI) for feature selection and Relative Weight K-Nearest Neighbor (RWKNN) for unbalanced text categorization method. CGI can select representative features for each category in the unbalanced text. RWKNN can overcome the problems of unbalanced text categorization. The experiment shows that the accuracy of CGI-RWKNN is better than CGI-KNN, at least 5% improved. CGI-RWKNN can select representative features for each category, show better results and stable unbalanced text categorization.

Item Type: Conference or Workshop Item (Paper)
Subjects: T Technology > T Technology (General)
Divisions: Faculty of Engineering > Department of Informatic
Depositing User: Monica Widiasri 61151
Date Deposited: 04 Sep 2014 01:50
Last Modified: 04 Sep 2014 01:50
URI: http://repository.ubaya.ac.id/id/eprint/20524

Actions (login required)

View Item View Item