ANALISIS SENTIMEN PUBLIK TERHADAP PRODUK REKOMENDASI INFLUENCER MENGGUNAKAN NAIVE BAYES: Studi Kasus Tasya Farasya

Garda Khoerunnisa, - and Syti Sarah Maesaroh, - and M. Rizki Nugraha, - (2025) ANALISIS SENTIMEN PUBLIK TERHADAP PRODUK REKOMENDASI INFLUENCER MENGGUNAKAN NAIVE BAYES: Studi Kasus Tasya Farasya. S1 thesis, Universitas Pendidikan Indonesia.

Abstract

Media sosial telah mengubah pola konsumsi informasi dan menjadikan influencer sebagai rujukan utama, khususnya dalam industri kecantikan. Penelitian ini bertujuan untuk menelaah persepsi publik terhadap produk Wardah yang berlabel Tasya Farasya Approved, serta menguji efektivitas metode Semi-Supervised Learning (SSL) dalam analisis sentimen komentar di TikTok. Sebanyak 2.716 komentar berhasil dikumpulkan melalui web scraping dari video milik Tasya Farasya dan akun resmi Wardah selama tahun 2024–2025. Setelah melalui tahap pelabelan dan preprocessing, diperoleh 878 komentar yang kemudian dibagi menjadi 80% data berlabel yang digunakan sebagai data latih dan data uji, serta 20% data tidak berlabel yang digunakan dalam proses pseudo-labeling. Seluruh data direpresentasikan menggunakan metode TF-IDF dan diklasifikasikan menggunakan algoritma Naive Bayes. Model divalidasi menggunakan 5-fold-cross validation dan diseimbangkan dengan teknik SMOTE. SSL diterapkan melalui metode pseudo-labeling terhadap data tidak berlabel dengan tingkat kepercayaan ≥ 80%. Hasil akhir menunjukkan distribusi sentimen terdiri dari 425 komentar positif, 305 netral, dan 148 negatif. Sentimen positif didominasi oleh apresiasi terhadap produk micellar water. Pengaruh Tasya Farasya sebagai influencer turut memperkuat persepsi positif publik. Penerapan SSL meningkatkan akurasi model dari 0,7565 menjadi 0,7733, disertai peningkatan precision, recall, dan F1-score. Social media has transformed information consumption patterns and positioned influencers as key references, particularly in the beauty industry. This study aims to examine public perception of Wardah products labeled as “Tasya Farasya Approved” and to evaluate the effectiveness of the Semi-Supervised Learning (SSL) method in sentiment analysis of TikTok comments. A total of 2,716 comments were collected through web scraping from videos posted by Tasya Farasya and Wardah’s official account during 2024–2025. After the labeling and preprocessing stages, 878 comments were obtained and split into 80% labeled data used for training and testing, and 20% unlabeled data used in the pseudo-labeling process. All data were represented using the TF-IDF method and classified using the Naive Bayes algorithm. The model was validated using 5-fold cross-validation and balanced using the SMOTE technique. SSL was applied through pseudo-labeling on unlabeled data with a confidence level of ≥ 80%. The final results show a sentiment distribution of 425 positive, 305 neutral, and 148 negative comments. Positive sentiment was predominantly driven by appreciation for the micellar water product. Tasya Farasya’s influence as an influencer further reinforced the public’s positive perception. The implementation of SSL improved the model’s accuracy from 0.7565 to 0.7733, along with increases in precision, recall, and F1-score.

	Text S_BIDI_2107906_Title.pdf Download (1MB)
	Text S_BIDI_2107906_Chapter1.pdf Download (380kB)
	Text S_BIDI_2107906_Chapter2.pdf Restricted to Staf Perpustakaan Download (827kB)
	Text S_BIDI_2107906_Chapter3.pdf Download (713kB)
	Text S_BIDI_2107906_Chapter4.pdf Restricted to Staf Perpustakaan Download (4MB)
	Text S_BIDI_2107906_Chapter5.pdf Download (233kB)
	Text S_BIDI_2107906_Appendix.pdf Restricted to Staf Perpustakaan Download (1MB)

Item Type:	Thesis (S1)
Additional Information:	https://scholar.google.com/citations?view_op=list_works&hl=en&user=MzMRMMkAAAAJ ID SINTA Dosen Pembimbing: Syti Sarah Maesaroh: 6681118 M. Rizki Nugraha: 6770726
Uncontrolled Keywords:	Analisis Sentimen, Semi-Supervised Learning, TikTok, Naive Bayes, Tasya Farasya Approved Sentiment Analysis, Semi-Supervised Learning, TikTok, Naive Bayes, Tasya Farasya Approved
Subjects:	L Education > L Education (General)
Divisions:	UPI Kampus Tasikmalaya > S1 Bisnis Digital
Depositing User:	Garda Khoerunnisa
Date Deposited:	26 Nov 2025 06:47
Last Modified:	28 Nov 2025 09:51
URI:	http://repository.upi.edu/id/eprint/141171

Actions (login required)

View Item