Skripsi

PENERAPAN METODE LONG-SHORT TERM MEMORY DAN FASTTEXT PADA KLASIFIKASI UJARAN KEBENCIAN

---

Hate speech on digital platforms poses a serious threat to social cohesion as it can lead to discrimination, conflict, and violence. This highlights the importance of developing automated detection systems to mitigate its impact and create safer digital spaces. To address this challenge, this study aims to develop a hate speech classification model in the Indonesian language using the Long Short-Term Memory (LSTM) method combined with the Fasttext Word Embedding approach. Fasttext was chosen for its ability to capture sub-word meanings and handle spelling variations, while LSTM excels in processing sequential data and understanding textual context comprehensively. The dataset used in this study comprises 2,980 social media comments categorized into two classes: hate speech and non-hate speech. The best-performing model, configured with 128 neurons in the LSTM layer, a dropout rate of 0.7, a learning rate of 0.1, a batch size of 64, and trained for 30 epochs, achieved an accuracy of 92.24%, a precision of 92.58%, a recall of 92.58%, and an F1-score of 92.58%. These results demonstrate that the developed model exhibits strong performance in classifying comments with high accuracy and robust generalization capabilities.

---

Availability

Detail Information

Series Title

-

Call Number

T1629192024

Publisher

Indralaya : Prodi Teknik Informatika, Fakultas Ilmu Komputer Universitas Sriwijaya., 2024

Collation

x, 83 hlm,; ilus.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

004.071 07

Content Type

Text

Media Type

unmediated

Carrier Type

other (computer)

Edition

-

Subject(s)

Pendidikan Teknik Informatika
Prodi Teknik Informatika

Specific Detail Info

-

Statement of Responsibility

MURZ

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment