Ramadhani, Oktavia (2020) Klasifikasi Multi-Label Berita Online Menggunakan Problem Transformation Dengan Metode K-Nearest Neighbor. Other thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
06211640000067-Undergraduate_Thesis.pdf Download (3MB) | Preview |
Abstract
Detik.com merupakan salah satu portal berita online paling populer di Indonesia. Portal berita online ini memiliki banyak kategori utama. Berita yang tersaji tidak selalu memuat satu kategori utama saja, akan tetapi dianggap hanya masuk dalam satu kategori utama saja. Permasalahan tersebut dapat diselesaikan dengan melakukan klasifikasi multi-label. Oleh karena itu, pada penelitian ini dilakukan klasifikasi multi-label menggunakan tiga metode problem transformation, yakni Binary Relevance, Label Powerset, dan Classifier Chain dengan metode klasifikasi dasar yakni K-Nearest Neighbor. Data yang digunakan dalam penelitian ini adalah data judul berita pada portal berita online detik.com yang terkategori secara multi-label dalam enam kategori, yakni detikFinance, detikOto, detikHot, detikInet, detikTravel, dan detikNews. Berdasarkan hasil analisis, didapatkan bahwa pada kasus ini metode problem transformation terbaik menggunakan metode klasifikasi dasar K-Nearest Neighbor yakni metode Binary Relevance. Hal tersebut didasari atas nilai hamming loss yang dihasilkan oleh metode Binary Relevance lebih kecil dibandingkan dengan metode Label Powerset dan Classifier Chain.
===========================================================
Detik.com is one of the most popular online news portals in Indonesia. Detik.com has various main categories. Every news presented in this news portal does not always contain only one main category, but it is classified only into one main category. This kind of problem can be solved by doing multi-label classification to classify news into one or more labels. Therefore, in this study, a multi-label classification was carried out using three problem transformation methods, which is Binary Relevance, Label Powerset, and Classifier Chain, with the K-Nearest Neighbor as the base classifier. The data used in this study is the news headlines data obtained through the online news portal detik.com. The news headlines are categorized as multi-label in six categories (detikFinance, detikOto, detikHot, detikInet, detikTravel, and detikNews). Based on the analysis results, it can be concluded that in this case, the best problem transformation method with K-Nearest Neighbor as the base classifier is the Binary Relevance method. It is based on the value of the hamming loss given by Binary Relevance method is smaller than the Label Powerset and Classifier Chain methods.
| Item Type: | Thesis (Other) |
|---|---|
| Additional Information: | RSSt 519.53 Ram k-1 • Ramadhani, Oktavia |
| Uncontrolled Keywords: | Hamming Loss, K-Nearest Neighbor, Klasifikasi Multi-Label, Problem Transformation, Hamming Loss, K-Nearest Neighbor, Multi-Label Classification, Problem Transformation |
| Subjects: | H Social Sciences > HD Industries. Land use. Labor > HD108 Classification (Theory. Method. Relation to other subjects ) |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis |
| Depositing User: | Oktavia Ramadhani |
| Date Deposited: | 03 Sep 2020 05:45 |
| Last Modified: | 21 Dec 2023 03:10 |
| URI: | http://repository.its.ac.id/id/eprint/81148 |
Actions (login required)
 |
View Item |