Klasifikasi Emosi Komentar YouTube pada Lagu Karya Idgitaf Mengggunakan Metode Cost-Sensitive Support Vector Machine (CS-SVM)

Sitanggang, Desi Daomara (2026) Klasifikasi Emosi Komentar YouTube pada Lagu Karya Idgitaf Mengggunakan Metode Cost-Sensitive Support Vector Machine (CS-SVM). Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (COVER) 22083010027-cover.pdf Download (1MB)
	Text (BAB I) 22083010027-bab1.pdf Download (248kB)
	Text (BAB II) 22083010027-bab2.pdf Restricted to Repository staff only until 14 September 2028. Download (436kB)
	Text (BAB III) 22083010027-bab3.pdf Restricted to Repository staff only until 14 September 2028. Download (4MB)
	Text (BAB IV) 22083010027-bab4.pdf Restricted to Repository staff only until 14 September 2028. Download (4MB)
	Text (BAB V) 22083010027-bab5.pdf Download (223kB)
	Text (DAFTAR PUSTAKA) 22083010027-daftarpustaka.pdf Download (181kB)
	Text (LAMPIRAN) 22083010027-lampiran.pdf Restricted to Repository staff only Download (219kB)

Abstract

YouTube is a social media platform that enables users to express their opinions and emotions through the comment section. Comments on the song “Takut” and “Satu Satu” by Idgitaf contain various emotional expressions that can be analyzed using Natural Language Processing (NLP) techniques. However, emotion classification in text data faces the challenge of class imbalance, which can affect the model’s ability to recognize emotion categories with fewer samples. This study aims to classify emotions in YouTube comments on the song “Takut” by Idgitaf using the Cost-Sensitive Support Vector Machine (CS-SVM) method with Word2Vec feature extraction, while also comparing the performance of four kernel types: Linear, Polynomial, Radial Basis Function (RBF), and Sigmoid. The research data were collected through crawling using the YouTube Data API v3, resulting in 20,644 comments. The research stages included preprocessing steps consisting of cleaning, case folding, tokenizing, normalization, filtering, and stemming, followed by emotion labeling using the NRC Emotion Lexicon. The labeled data were then represented as vectors using Word2Vec and classified using CS-SVM with class weighting to address data imbalance. Model evaluation was conducted using accuracy, precision, recall, and F1-score. The results showed that trust was the most dominant emotion category, followed by fear and anticipation, while surprise was the category with the smallest amount of data. The comparison of the four kernels indicated that the RBF kernel achieved the best performance, with an accuracy of 70.87%, compared to the Linear kernel at 70.63%, the Polynomial kernel at 68.43%, and the Sigmoid kernel at 63.98%. Visualization analysis showed overlap among emotion categories in the Word2Vec feature space, while the WordCloud results indicated that the fear, sadness, and joy categories had dominant words that were relatively consistent with their emotional characteristics. The CS-SVM model with the RBF kernel was then implemented in a Flask-based Graphical User Interface (GUI) to classify emotions in new comments interactively.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Saputra, Wahyu Syaifullah Jauharis	NIDN0725088601	wahyu.s.j.saputra.if@upnjatim.ac.id
Thesis advisor	Nasrudin, Muhammad	NUPTK4241774675130323	nasrudin.fasilkom@upnjatim.ac.id

Subjects:

H Social Sciences > HG Finance > HG1709 Data processing
Q Science > QA Mathematics
Q Science > QA Mathematics > QA76.87 Neural computers

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

Desi Daomara Sitanggang

Date Deposited:

15 Sep 2026 03:46

Last Modified:

15 Sep 2026 03:54

URI:

https://repository.upnjatim.ac.id/id/eprint/60172