<![CDATA[Forest and land fires in Central Kalimantan are environmental problems influenced by natural factors and human activities, resulting in complex causal relationship patterns. This study aims to analyze wildfire cause patterns in Central Kalimantan using an IndoBERT-based NLP and AI approach integrated with causal pattern mining on online news articles. Data were collected through web scraping from six local and national news portals using nine Google search queries related to wildfires, resulting in 436 relevant articles as the main corpus. The methodological stages included text preprocessing, semantic representation using a 768-dimensional IndoBERT transformer model, topic discovery using BERTopic, and causal pattern analysis through co-occurrence analysis and contextual relation mining. The topic discovery results identified 10 main topics, with the topic “extreme dry season” dominating with 104 data points, followed by “hot weather and drought” (54 data points). Anthropogenic factors such as intentional land burning, land clearing activities, and human negligence were also identified as significant causes. Contextual relation mining results showed that the words “land” (1,473 occurrences), “smoke” (938), “dry season” (334), and “peatland” (288) were the most dominant causal contexts, while co-occurrence analysis generated 37,581 word pairs forming a causal network. Future studies are recommended to integrate spatial data, expand the dataset, and implement knowledge graphs to support real-time wildfire disaster intelligence systems.]]>
