Deteksi Cyberbullying pada Media Sosial X Menggunakan IndoBERTweet-Bi-GRU

Nafiah, Fajria Ulumin (2026) Deteksi Cyberbullying pada Media Sosial X Menggunakan IndoBERTweet-Bi-GRU. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) Cover.pdf Download (646kB)
	Text (Bab 1) Bab 1.pdf Download (37kB)
	Text (Bab 2) Bab 2.pdf Restricted to Repository staff only until 30 October 2029. Download (439kB)
	Text (Bab 3) Bab 3.pdf Restricted to Repository staff only until 30 October 2029. Download (361kB)
	Text (Bab 4) Bab 4.pdf Restricted to Repository staff only until 30 October 2029. Download (770kB)
	Text (Bab 5) Bab 5.pdf Download (103kB)
	Text (Daftar Pustaka) Daftar Pustaka.pdf Download (142kB)
	Text (Lampiran) Lampiran.pdf Restricted to Repository staff only until 30 October 2029. Download (197kB)

Abstract

Social media users in Indonesia reached 143 million in early 2025, accounting for approximately 50.2% of the total population. This figure increased from 139 million users in the previous year, indicating that social media has become an increasingly important part of people's lives. However, the growing use of social media may also create new problems, one of which is cyberbullying. According to a report by UNICEF Indonesia, 45% of adolescents and young adults aged 14–24 years have experienced cyberbullying. Twitter (now X) is one of the platforms vulnerable to such behavior due to its characteristics that allow users to express opinions freely and impulsively, creating opportunities for negative comments, hate speech, and bullying. This study aims to develop a cyberbullying detection model for Indonesian-language text using a Natural Language Processing (NLP) approach. The proposed model combines IndoBERTweet, a pre-trained transformer-based model optimized for Indonesian social media text, with a Bi-directional Gated Recurrent Unit (Bi-GRU) architecture that is effective in capturing sequential context. Focal Loss is applied to address class imbalance during model training. IndoBERTweet generates contextual text representations, which are subsequently processed by Bi-GRU to capture information in both forward and backward directions before being classified into cyberbullying categories. The best performance was achieved by combining Focal Loss with a modified four-class label configuration, namely Rude and Vulgar Words, Sexual Harassment, Body Shaming and Hate Speech, and Non-Cyberbullying, achieving an accuracy of 0.93, precision of 0.90, recall of 0.90, and an F1-score of 0.90.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Damaliana, Aviolla Terza	NIDN0002089402	aviolla.terza.sada@upnjatim.ac.id
Thesis advisor	Hindrayani, Kartika Maulida	NIDN0009099205	kartika.maulida.ds@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics
Q Science > QA Mathematics > QA75 Electronic computers. Computer science

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

Fajria Ulumin Nafiah

Date Deposited:

30 Oct 2026 02:57

Last Modified:

30 Oct 2026 03:06

URI:

https://repository.upnjatim.ac.id/id/eprint/61199