<![CDATA[Twitter (now X) is a critical social media platform for disseminating information during crises. This study models disaster-related topics from Indonesian-language tweets using Latent Dirichlet Allocation (LDA). From a dataset of 8,718 tweets collected from official sources like BMKG and BNPB, we performed several preprocessing steps, including case folding, stop word removal, stemming, and normalization of slang and abbreviations. The optimal number of topics was determined using coherence scores, with the model achieving a peak coherence value of approximately 0.57. Keywords such as “banjir”, “kecelakaan”, “tanah longsor,” and others were used to collect data from Twitter accounts like "BMKG" (Meteorology, Climatology, and Geophysical Agency) and "BNPB" (National Disaster Management Agency). The results revealed that the most frequently discussed topics with high coherence values were “angin topan” “topan”, “virus corona”, “kecelakaan”, “tenggelam”, “badai”, “angin puting.” A word cloud was used to visualize these disaster-related topics.]]>
Copyrights © 2025
