Skripsi

KLASIFIKASI MULTILABEL KOMENTAR PADA TWITTER MENGGUNAKAN LONG SHORT TERM MEMORY

---

Penilaian

---

Twitter is one of the most popular social media platforms and users can write comments in the form of tweets freely without any restrictions. These tweets can contain various blasphemies and toxic comments. Toxic comments are comments that are rude, disrespectful, unreasonable, or even to the point of humiliating someone. They can cause serious problems on social media and some people will avoid engaging in unfair and unhealthy debates. Toxic comments can consist of several labels. This research aims to perform multilabel classification of comments. The method used is Long Short Term Memory and Word2Vec as word embedding. The data used amounted to 2,682 tweets which were then divided into 80% training data and 20% test data. After tuning the hyperparameters using random search, the best results were obtained for the LSTM model with a dropout configuration of 0.2, hidden unit 128, recurrent dropout in the LSTM layer 0.3, epochs 20, and batch size 64. Based on the research results, the average value of hamming loss is 0.138.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2307003696	T127461	T1274612023	Central Library (Reference)	Available but not for loan - Not for Loan

Detail Information

Series Title

-

Call Number

T1274612023

Publisher

Indralaya : Jurusan Teknik Informatika, Fakultas Ilmu Komputer Universitas Sriwijaya., 2023

Collation

xv, 77 hlm.; Ilus.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

006.754

Content Type

-

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Twitter
Jurusan Teknik Informatika

Specific Detail Info

-

Statement of Responsibility

FIRA

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment