Analisis Sentimen Komentar YouTube pada Video Terkait Insiden Pengemudi Ojek Online dan Anggota Brimob Menggunakan Algoritma Naive

Media sosial telah menjadi ruang ekspresi publik yang dinamis, di mana masyarakat menyampaikan opini terhadap berbagai peristiwa aktual. Penelitian ini bertujuan untuk menganalisis sentimen masyarakat terhadap insiden antara pengemudi ojek online dan anggota Brimob melalui komentar YouTube. Metode yang digunakan adalah algoritma Naive Bayes dengan pendekatan klasifikasi teks. Data dikumpulkan dari 10 video YouTube, menghasilkan 1.143 komentar, yang setelah dibersihkan menjadi 1.121 komentar. Fitur teks dibentuk menggunakan TF-IDF Vectorizer dan data dibagi menjadi data latih (896) dan data uji (225). Hasil klasifikasi menunjukkan distribusi sentimen: netral (48.7%), negatif (46.7%), dan positif (4.5%). Evaluasi model menghasilkan akurasi 67%, precision 46%, recall 47%, dan F1-score 45%. Temuan ini menunjukkan bahwa mayoritas komentar bersifat netral dan negatif, serta bahwa Naive Bayes cukup efektif dalam klasifikasi opini publik meskipun memiliki keterbatasan dalam menangani data minoritas. Kata Kunci: Analisis Sentimen, Komentar YouTube, Naive Bayes, Text Mining, Opini Publik.   Social media has become a dynamic space for public expression, where individuals share opinions on various current events. This study aims to analyze public sentiment regarding the incident between an online motorcycle taxi driver and a member of Brimob through YouTube comments. The method employed is the Naive Bayes algorithm with a text classification approach. Data was collected from 10 YouTube videos, yielding 1,143 comments, which were cleaned down to 1,121 comments for analysis. Text features were constructed using the TF-IDF Vectorizer, and the dataset was split into training data (896 comments) and test data (225 comments). The classification results show sentiment distribution as follows: neutral (48.7%), negative (46.7%), and positive (4.5%). Model evaluation produced an accuracy of 67%, precision of 46%, recall of 47%, and F1-score of 45%. These findings indicate that most comments are neutral and negative, and that Naive Bayes is reasonably effective in classifying public opinion, although it faces challenges in handling minority classes such as positive sentiment. Keywords: Sentiment Analysis, YouTube Comments, Naive Bayes, Text Mining, Public Opinion.