Performance comparison of machine learning algorithms for forest fire detection in Peninsular Malaysia
Yee Jian Chew, Shih Yin Ooi, Ying Han Pang · 2025
This paper presents a pilot study evaluating 13 machine learning classifiers for detecting forest fires in Peninsular Malaysia using a forest fire inventory dataset generated through the Google Earth Engine (GEE) framework, developed in our previous work. The experimental results demonstrate the suitability of machine learning techniques for forest fire detection in the region. Among the classifiers tested, tree-based models outperformed others, with Random Forest achieving the highest recall of 99.7876%, followed closely by Gradient Boosting with a recall of 99.7345%. These findings suggest that tree-based classifiers are particularly well-suited for forest fire detection tasks. Future work is recommended to focus on refining or enhancing these models to further improve detection performance.