Topic Modeling of Natural Disaster in Indonesia Using NMF

Natural disasters pose significant challenges in Indonesia, a geographically vulnerable nation prone to earthquakes, floods, and volcanic eruptions. This research applies the Non-Negative Matrix Factorization (NMF) technique for topic modeling to comprehensively analyze text datasets related to natural disasters in Indonesia. The study explores the impact of varying the number of topics (5, 10, and 15) on the effectiveness of NMF in representing the underlying structure of the text data. The findings reveal that the choice of the number of topics significantly influences the granularity and richness of text data representation. A lower reconstruction error is achieved with 15 topics, indicating its effectiveness in capturing the intricate relationships between natural disasters and various domains, including education, agriculture, sports, and regional events. The 5 major themes identified encompass “Floods and Landslides,” “Earthquakes,” “Normal Weather,” and “Other,” providing valuable insights into disaster management and mitigation efforts. Additionally, the analysis with 10 topics expands the scope to encompass legal issues, environmental concerns, and political aspects, reflecting the complex interplay between disasters and broader societal issues. This research underscores the importance of adopting nuanced topic modeling approaches to understand the multifaceted nature of natural disasters and their implications for various sectors. Moreover, it highlights the significance of the reconstruction error score as a crucial metric for evaluating NMF model performance in text data analysis. The study contributes to enhancing disaster preparedness and addressing environmental challenges in Indonesia, offering valuable lessons for disaster management and text analysis methodologies, as well as providing information for more targeted mitigation planners.