DETEKSI GEJALA STRES PADA POSTINGAN TEKS SOSIAL MEDIA X BERBAHASA INDONESIA MENGGUNAKAN MODEL BERBASIS DEBERTAV3

Fikri Arif Rahman, - (2025) DETEKSI GEJALA STRES PADA POSTINGAN TEKS SOSIAL MEDIA X BERBAHASA INDONESIA MENGGUNAKAN MODEL BERBASIS DEBERTAV3. S1 thesis, Universitas Pendidikan Indonesia.

Abstract

Penggunaan media sosial dengan intensitas tinggi di masyarakat berpotensi menimbulkan dampak negatif terhadap kesehatan mental, salah satunya berupa gejala stres. Namun, kondisi ini kerap tidak teridentifikasi karena individu jarang mengekspresikan perasaan secara langsung. Stres yang tidak terdeteksi sejak dini dapat berkembang menjadi gangguan psikologis yang lebih serius, sehingga diperlukan upaya sistematis untuk mendukung deteksi dini berbasis teknologi. Penelitian ini bertujuan mengembangkan metode deteksi gejala stres melalui analisis teks tidak terstruktur pada cuitan media sosial X berbahasa Indonesia. Metodologi yang digunakan mencakup tahapan pengumpulan dataset berlabel stres dan tidak stres, pra-pemrosesan teks, pelatihan model DeBERTaV3, serta evaluasi kinerja menggunakan confusion matrix. Dataset terdiri dari 2.200 data latih dan 550 data uji. Hasil penelitian menunjukkan bahwa model DeBERTaV3 mampu mendeteksi gejala stres dengan akurasi 81%, presisi 82%, recall 81%, dan F1-score 78%. Temuan ini menegaskan efektivitas DeBERTaV3 dalam menganalisis teks media sosial yang sarat bahasa informal, singkatan, dan emotikon. --------- The intensive use of social media within society has the potential to generate negative impacts on mental health, one of which is stress symptoms. However, such conditions often remain unidentified as individuals rarely express their feelings explicitly. Undetected stress can escalate into more serious psychological disorders, thereby necessitating systematic efforts to support early detection through technology-based approaches. This study aims to develop a method for detecting stress symptoms through the analysis of unstructured text from Indonesian-language posts on social media platform X. The methodology encompasses several stages, including the collection of labeled datasets (stress and non-stress), text preprocessing, training of the DeBERTaV3 model, and performance evaluation using a confusion matrix. The dataset consisted of 2,200 training data and 550 testing data. The results indicate that the DeBERTaV3 model achieved 81% accuracy, 82% precision, 81% recall, and a 78% F1-score in stress detection. These findings demonstrate the effectiveness of DeBERTaV3 in analyzing social media texts that contain informal language, abbreviations, and emoticons.

	Text S_RPL_2007162_Title.pdf Download (3MB)
	Text S_RPL_2007162_Chapter1.pdf Download (1MB)
	Text S_RPL_2007162_Chapter2.pdf Restricted to Staf Perpustakaan Download (5MB) | Request a copy
	Text S_RPL_2007162_Chapter3.pdf Download (1MB)
	Text S_RPL_2007162_Chapter4.pdf Restricted to Staf Perpustakaan Download (4MB) | Request a copy
	Text S_RPL_2007162_Chapter5.pdf Download (293kB)
	Text S_RPL_2007162_Appendix.pdf Restricted to Staf Perpustakaan Download (1MB) | Request a copy

Item Type:	Thesis (S1)
Additional Information:	scholar.google.com/citations?view_op=new_profile&hl=en&authuser=1 SINTA ID : Raditya Muhammad 6682222 Indira Syawanodya 6681751
Uncontrolled Keywords:	Media sosial, Gejala stres, Deteksi dini, Natural Language Processing, DeBERTaV3, Social media, Stress symptoms, Early detection
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software
Divisions:	UPI Kampus cibiru > S1 Rekayasa Perangkaat Lunak
Depositing User:	Fikri Arif Rahman
Date Deposited:	18 Sep 2025 04:23
Last Modified:	18 Sep 2025 04:23
URI:	http://repository.upi.edu/id/eprint/137567

Actions (login required)

View Item