Association on Supervised Term Weighting Method for Classification on Data Twitter
Imroatul Khuluqi Izzah1, Abba Suganda Girsang2
1Imroatul Khuluqi Izzah, Computer Science Department, BINUS Graduate Program – Master of Computer Science, Bina Nusantara University, Jakarta, Indonesia.
2Abba Suganda Girsang, Computer Science Department, BINUS Graduate Program, Bina Nusantara University, Jakarta, Indonesia.
Manuscript received on February 02, 2020. | Revised Manuscript received on February 10, 2020. | Manuscript published on March 30, 2020. | PP: 859-863 | Volume-8 Issue-6, March 2020. | Retrieval Number: F6975038620/2020©BEIESP | DOI: 10.35940/ijrte.F6975.038620
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Term weighting is a preprocessing phase that has an important role in the text classification by giving the appropriate weight for each term in all documents. In previous research, many supervised term weighting methods have been introduced, but most of the supervised term weighting only considers the distribution of terms in the two classes so that it is not optimal for the multi-class classification. This paper introduces a new supervised weighting with association concept to optimize term weighting distributions in multi-class cases by considering terms that exist in each class and paying attention to the number of terms in the document belonging to the class, also considering the relationship pattern between one or more items with association concept in a dataset to measure the strength of terms in a class by using confidence values. The dataset used are the data twitter taken from the PR FM twitter account. The proposed supervised term weighting method implemented with SVM classifier can outperform unsupervised weighting schemes such as TF-IDF with the average accuracy 81.704%.
Keywords: Term-Weighting, Classification, Twitter, Association, Confidence.
Scope of the Article: Classification.