<![CDATA[Abstract TikTok has become one of the social media platforms widely used by the public to express opinions regarding the flood disaster in Sumatra. This study aims to analyze discussion topics and classify the sentiment of TikTok comments using Aspect-Based Sentiment Analysis (ABSA) by integrating Latent Dirichlet Allocation (LDA) and Extreme Gradient Boosting (XGBoost). The research data consisted of 10,000 comments from 12 TikTok videos collected from November 26, 2025, to January 28, 2026. The research stages included text preprocessing, LDA topic modeling, TF-IDF feature extraction, and sentiment classification using XGBoost. The preprocessing process resulted in 5,555 comments suitable for analysis. LDA identified three main aspects with a coherence score of 0.4592: Community Empathy and Reactions, Flood Conditions and Management, and Prayers and Hopes for Victims. Sentiment was dominated by the neutral class, with 2,372 comments. XGBoost testing showed that the 70:30 data split achieved the best performance, with an accuracy of 90.82% and precision, recall, and F1-score of 0.91. These results demonstrate that the combination of LDA and XGBoost is effective for aspect-based sentiment analysis. Keywords: Aspect-Based Sentiment Analysis; TikTok; Sumatra Flood; Latent Dirichlet Allocation; XGBoost Abstrak TikTok menjadi salah satu media sosial yang banyak digunakan masyarakat untuk menyampaikan opini mengenai bencana banjir di Sumatera. Penelitian ini bertujuan menganalisis topik pembahasan dan mengklasifikasikan sentimen komentar TikTok menggunakan Aspect-Based Sentiment Analysis (ABSA) dengan integrasi Latent Dirichlet Allocation (LDA) dan Extreme Gradient Boosting (XGBoost). Data penelitian berupa 10.000 komentar dari 12 video TikTok pada 26 November 2025 hingga 28 Januari 2026. Tahapan penelitian meliputi pra-pemrosesan teks, pemodelan topik LDA, ekstraksi fitur TF-IDF, dan klasifikasi sentimen menggunakan XGBoost. Hasil pra-pemrosesan menghasilkan 5.555 komentar layak analisis. LDA menghasilkan tiga aspek utama dengan coherence score 0,4592, yaitu Empati dan Reaksi Masyarakat, Kondisi dan Penanganan Banjir, serta Doa dan Harapan bagi Korban. Sentimen didominasi netral sebanyak 2.372 komentar. Pengujian XGBoost menunjukkan rasio data 70:30 memberikan performa terbaik dengan akurasi 90,82% serta precision, recall, dan F1-score sebesar 0,91. Hasil tersebut menunjukkan kombinasi LDA dan XGBoost efektif dalam analisis sentimen berbasis aspek. Kata Kunci: Aspect-Based Sentiment Analysis; TikTok; Banjir Sumatera; Latent Dirichlet Allocation; XGBoost]]>
