Article Classification Using Convolutional Neural Network (CNN) And Chi-Square Feature Selection

  • I Gede Laksmana Yudha Universitas Udayana
  • Ngurah Agus Sanjaya ER Universitas Udayana
  • Anak Agung Istri Ngurah Eka Karyawati Universitas Udayana
  • Ida Bagus Gede Dwidasmara Universitas Udayana
  • I Gusti Ngurah Anom Cahyadi Putra Universitas Udayana
  • Ida Bagus Made Mahendra Universitas Udayana

Abstract

News articles are reports or information about events that are actual, reliable, and based on facts or reality. The increase in internet users has resulted in the growth of the amount of available information increasing rapidly. Easy internet access makes many types of Indonesian news articles published digitally. With a very large number of news articles, it will be easier to find a news article if the news has been organized and has been grouped according to its respective categories. Text classification is a problem that aims to determine the topic or theme of a document. In achieving this goal, the classification process forms a model that can distinguish data into different classes based on certain rules. The method used to build the model is Convolutional Neural Network (CNN) with Chi-Square feature selection. News articles are divided into six classes, namely news, technology, football, health, lifestyle, and automotive. In this study, the best CNN model was obtained with the number of filters used was 200 and the feature selection being 40%. The test results on the test data provide an accuracy value of 96,074%, precision of 96,079%, recall of 96,074%, and an f-1 score of 96,070%.

Downloads

Download data is not yet available.
Published
2022-07-11
How to Cite
YUDHA, I Gede Laksmana et al. Article Classification Using Convolutional Neural Network (CNN) And Chi-Square Feature Selection. JELIKU (Jurnal Elektronik Ilmu Komputer Udayana), [S.l.], v. 11, n. 3, p. 529-538, july 2022. ISSN 2654-5101. Available at: <https://ojs.unud.ac.id/index.php/jlk/article/view/88589>. Date accessed: 30 may 2024. doi: https://doi.org/10.24843/JLK.2023.v11.i03.p08.