<![CDATA[Perkembangan teknologi komunikasi digital telah melahirkan inovasi baru seperti embedded SIM (eSIM), yang menawarkan kemudahan dalam pengelolaan identitas pelanggan seluler tanpa kartu fisik. Seiring meningkatnya adopsi teknologi ini, YouTube menjadi salah satu media diskusi publik yang ramai membahas eSIM melalui kolom komentar. Penelitian ini bertujuan untuk menganalisis sentimen masyarakat terhadap layanan eSIM berdasarkan komentar-komentar di video YouTube. Dengan menggunakan algoritma klasifikasi Naive Bayes dan Random Forest, sebanyak 324 komentar dikategorikan menjadi opini positif dan negatif. Proses penelitian mencakup tahapan pengumpulan data melalui teknik scraping, preprocessing teks, serta evaluasi model menggunakan metrik akurasi, presisi, recall, dan AUC. Hasil menunjukkan bahwa algoritma Naive Bayes mampu mencapai akurasi sebesar 98,52% dengan presisi tinggi terutama pada kelas negatif. Sementara itu, Random Forest menghasilkan akurasi lebih tinggi sebesar 99,69% dengan nilai AUC sempurna sebesar 1.000, mencerminkan performa optimal dalam membedakan sentimen komentar. Temuan ini menegaskan bahwa kedua algoritma efektif dalam klasifikasi sentimen teks, dengan Random Forest menunjukkan keunggulan performa. Penelitian ini dapat menjadi referensi bagi pengembangan analisis opini publik secara digital serta pemanfaatan machine learning dalam pemrosesan bahasa alami. Kata kunci: Analisis Sentimen, Youtube, eSIM Abstract (10pt, italic, tebal, dan ditengah) The advancement of digital communication technologies has introduced new innovations such as the embedded SIM (eSIM), which provides users with flexibility in managing their mobile identity without the need for a physical SIM card. As the adoption of this technology increases, YouTube has become a prominent platform where public discussions regarding eSIM occur through comment sections. This study aims to analyze public sentiment towards eSIM services based on comments posted on YouTube videos. Utilizing the Naive Bayes and Random Forest classification algorithms, a total of 324 comments were categorized into positive and negative sentiments. The research process involved data collection through web scraping, text preprocessing, and model evaluation using metrics such as accuracy, precision, recall, and AUC. The results show that the Naive Bayes algorithm achieved an accuracy of 98.52%, with particularly high precision for the negative class. Meanwhile, the Random Forest algorithm yielded even higher accuracy at 99.69%, with a perfect AUC score of 1.000, indicating outstanding performance in distinguishing between sentiment classes. These findings affirm the effectiveness of both algorithms in sentiment text classification, with Random Forest demonstrating superior performance. This research contributes as a reference for further applications of public opinion analysis in digital media and the implementation of machine learning in natural language processing.Keywords: Sentiment Analysis, Youtube, eSIM]]>
