Dirgahayu Marganingsih, Hardian Oktavianto, Ginanjar Abdurrahman

Analisis Sentimen Komentar Youtube Masterchef Indonesia Menggunakan Algoritma Support Vector Machine dan Gaussian Naïve Bayes

Introduction

Analisis sentimen komentar youtube masterchef indonesia menggunakan algoritma support vector machine dan gaussian naïve bayes. Analisis sentimen komentar YouTube Masterchef Indonesia menggunakan SVM & Gaussian Naïve Bayes. Pelajari pro-kontra grand final & kinerja algoritma klasifikasi sentimen positif/negatif.

Abstract

Platform media sosial menjadi utama untuk berbagi informasi, termasuk platform berbasis video sharing seperti Youtube, yang menyediakan konten edukasi, informasi, dan hiburan bagi masyarakat. Salah satu konten yang banyak diminati adalah acara memasak yang diadaptasi dari Inggris, yaitu Masterchef Indonesia. Yang berjalan selama 11 musim dari tahun 2011 sampai 2023 dengan 260 episode. Namun, pada musim ke 11 episode grand final terdapat pro dan kontra tentang kemanangan peserta yang membuat publik terprovokasi dengan kehadiran calon presiden Ganjar Pranowo. Melalui fitur komentar, pengguna dapat menyampaikan pendapat dan tanggapan, yang kemudian dapat diinterpretasikan untuk memahami pandangan masyarakat. Penelitian ini bertujuan untuk menganalisis komentar yang bernilai positif dan negatif pada channel Masterchef Indonesia dengan algoritma Support Vector Machine dan Gaussian Naïve Bayes. Sebelum diterapkan teknik oversampling untuk penyeimbangan data, algoritma Support Vector Machine menghasilkan akurasi 82%, presisi 88%, dan recall 72%, sedangkan Gaussian Naïve Bayes menghasilkan akurasi 65%, presisi 52%, dan recall 81%. Setelah teknik oversampling diterapkan, performa Support Vector Machine meningkat dengan akurasi 85%, presisi 84%, dan recall 89%, sementara Gaussian Naïve Bayes memperoleh akurasi 72%, presisi 68%, dan recall 72%.

Review

The paper, "Analisis Sentimen Komentar Youtube Masterchef Indonesia Menggunakan Algoritma Support Vector Machine dan Gaussian Naïve Bayes," addresses a timely and relevant topic: sentiment analysis on social media, specifically YouTube comments. The chosen context, the Masterchef Indonesia Season 11 grand final, is particularly interesting due to the documented public controversy and strong emotional responses, making it an excellent case study for sentiment classification. The authors aim to classify positive and negative sentiment using two established machine learning algorithms, Support Vector Machine (SVM) and Gaussian Naïve Bayes (GNB), providing insights into public opinion surrounding a significant cultural event. The methodology employed is sound, leveraging two widely recognized classification algorithms, SVM and GNB, which are suitable choices for text classification tasks. A significant strength of this study is the explicit application and evaluation of an oversampling technique to address potential class imbalance, a common issue in real-world sentiment datasets. The reported performance metrics (accuracy, precision, recall) both before and after oversampling clearly demonstrate the effectiveness of this approach, particularly for SVM, which showed a notable improvement in all metrics, especially recall. This detailed comparison of algorithm performance with and without data balancing provides valuable insights into robust model development for sentiment analysis. While the study presents clear results and a robust methodology, the abstract could benefit from briefly mentioning the specific features used for classification (e.g., TF-IDF, word embeddings) and the language processing steps involved, which are crucial details for reproducibility and understanding the model's foundation. Furthermore, a brief discussion on the limitations, such as the generalizability of the models to different contexts or the handling of nuanced sentiment (e.g., sarcasm, irony), would strengthen the abstract's completeness. Nevertheless, this research offers a valuable contribution to the field of sentiment analysis, particularly in the context of Indonesian social media discourse, demonstrating the efficacy of SVM and GNB, enhanced by oversampling, in classifying public sentiment on contentious issues.

Full Text

Comments

You need to be logged in to post a comment.