<![CDATA[Forest fires pose a significant challenge in Riau Province, Indonesia, especially in peatland areas. This study employs the Random Forest (RF) algorithm to analyze the confidence levels of hotspots, aiming to predict potential fire occurrences and improve fire management strategies. The research focuses on peatlands spanning 3.86 million ha, using key variables such as NDVI, surface temperature, and peat thickness derived from satellite data. The model achieved an average AUC of 0.732 and a classification accuracy of 70.3%, with medium-confidence hotspots demonstrating the best predictive performance (AUC: 0.707, F1-score: 0.804). However, the model struggled with low-confidence hotspots, reflecting challenges in distinguishing less prominent patterns in the data. Compared to other methods, RF demonstrates strong potential in handling complex environmental datasets, making it a valuable tool for hotspot prediction. This study contributes to understanding forest fire risks in peatlands and provides actionable insights for improving preparedness and mitigation efforts.]]>
Copyrights © 2025
