Ihsan Ghozi Zulfikar, - and Yudi Wibisono, - and Asep Wahyudin, - (2025) SISTEM AGREGASI BERITA DENGAN KLASIFIKASI, CLUSTERING DAN PERINGKASAN OTOMATIS. S1 thesis, Universitas Pendidikan Indonesia.
Abstract
Berita online telah menjadi sumber utama informasi bagi masyarakat Indonesia, namun banyaknya berita online sering kali membuat pembaca kesulitan menemukan berita yang sesuai minat. Penelitian ini bertujuan mengembangkan sistem agregasi berita berbasis web yang dapat melakukan klasifikasi, clustering dan peringkasan secara otomatis terhadap artikel berita online berbahasa Indonesia menggunakan kecerdasan buatan. Sistem ini dibangun menggunakan pendekatan pengembangan ADDIE dan implementasi perangkat lunak dengan metode Waterfall. Proses dimulai dari scraping artikel dari Kompas dan TribunNews pada Juni 2025, menghasilkan 37.187 artikel. Model klasifikasi yang digunakan yaitu BLSTM-2DCNN yang dilatih menggunakan dataset Indosum mencapai akurasi 86% dan F1-score 0.85. Clustering dilakukan menggunakan k-means berdasarkan kategori dan rentang waktu lima hari. Evaluasi menggunakan CH-Index menunjukkan nilai rata-rata 80.87. Untuk peringkasan, digunakan model BART yang di-fine-tuning, dengan ROUGE-1, ROUGE-2, dan ROUGE-L masing masing meningkat menjadi 0.6434, 0.5510, dan 0.6077. Sistem yang dibangun menyajikan artikel yang telah diklasifikasikan, dikelompokkan dan diringkas melalui antarmuka berbasis Flutter dan backend Flask. Online news has become the primary source of information for Indonesian society; however, the overwhelming number of news articles often makes it difficult for readers to find content relevant to their interests. This research aims to develop a web-based news aggregation system capable of automatically performing classification, clustering, and summarization of Indonesian online news articles using artificial intelligence. The system was designed using the ADDIE development framework and implemented with the Waterfall software development method. Data were collected by scraping articles from Kompas and TribunNews in June 2025, resulting in 37,187 articles. The classification model employed a BLSTM-2DCNN trained on the Indosum dataset, achieving an accuracy of 86% and an F1-score of 0.85. Clustering was performed using k-means based on news categories and a five-day time window, with evaluation using the CH-Index yielding an average score of 80.87. For summarization, a fine-tuned BART model was applied, producing ROUGE-1, ROUGE-2, and ROUGE-L scores of 0.6434, 0.5510, and 0.6077, respectively. The developed system presents classified, clustered, and summarized articles through a Flutter-based interface with a Flask backend, enabling readers to access information more efficiently and comprehensively.
![[thumbnail of S_KOM_2103303_Title.pdf]](http://repository.upi.edu/style/images/fileicons/text.png "S_KOM_2103303_Title.pdf") |
Text
S_KOM_2103303_Title.pdf Download (500kB) |
|
Text
S_KOM_2103303_Chapter1.pdf Download (324kB) |
|
Text
S_KOM_2103303_Chapter2.pdf Restricted to Staf Perpustakaan Download (726kB) | Request a copy |
|
Text
S_KOM_2103303_Chapter3.pdf Download (166kB) |
|
Text
S_KOM_2103303_Chapter4.pdf Restricted to Staf Perpustakaan Download (672kB) | Request a copy |
|
Text
S_KOM_2103303_Chapter5.pdf Download (103kB) |
| Item Type: | Thesis (S1) |
|---|---|
| Additional Information: | https://scholar.google.com/citations?user=l5lXpBUAAAAJ&hl=en& ID SINTA Dosen Pembimbing: Yudi Wibisono: 260167 Asep Wahyudin: 5991982 |
| Uncontrolled Keywords: | Agregasi Berita, Clustering, Klasifikasi, Peringkasan Classification, Clustering, Summarization, News Aggregation |
| Subjects: | L Education > L Education (General) Q Science > QA Mathematics > QA75 Electronic computers. Computer science |
| Divisions: | Fakultas Pendidikan Matematika dan Ilmu Pengetahuan Alam > Program Studi Ilmu Komputer |
| Depositing User: | Ihsan Ghozi Zulfikar |
| Date Deposited: | 08 Sep 2025 04:35 |
| Last Modified: | 08 Sep 2025 04:35 |
| URI: | http://repository.upi.edu/id/eprint/137961 |
Actions (login required)
 |
View Item |