Analisis Topik Komentar Publik Berdasarkan Sentimen pada Tayangan Ruang Guru Clash of Champions Menggunakan SVM dan LDA

Abstract

Ruangguru Clash of Champions is an educational competition program broadcast through the Ruangguru YouTube channel and application, featuring university students from various institutions. Although the program received positive public response, it also faced criticism. To understand public opinion effectively, this study analyzes the large volume of unstructured comments using topic modeling and sentiment analysis. This research adopts a quantitative approach by converting unstructured text data into numerical form through weighting, classification, and probabilistic modeling. The Latent Dirichlet Allocation (LDA) method was employed to identify the main discussion topics and assign a topic label to each comment, as LDA has proven effective in modeling short-text data on a large scale. Support Vector Machine (SVM) was used for sentiment classification into two categories: positive and negative, due to its high accuracy in sentiment analysis. The dataset consists of 95,100 comments collected from 12 videos on the Ruangguru YouTube channel using the YouTube API. The results show four dominant topics: winner eligibility, educational aspirations, support for Ruangguru, and episode airing schedule. Each comment was successfully labeled with one of these topics. Sentiment distribution analysis indicates that educational aspirations and support for Ruangguru were dominated by positive sentiments, while winner eligibility and episode schedule received mostly negative sentiments. Topic modeling evaluation yielded a highest coherence score of 0.5014, while the SVM model achieved 96% accuracy using a 90:10 training-testing ratio.