Improving Fire Type Classification in Bahrain Using Supervised Machine Learning Techniques

Fire breakouts are inevitable and can lead to the destruction of buildings, infrastructures, urban areas, and often result in severe medical complications for affected inhabitants. Bahrain is particularly prone to such incidents, due to the high population density, proximity to the equator, and naturally dry climate. The usage of Machine Learning (ML) for fire identification significantly increased the speed of classification while achieving increased reliability and accuracy. This study evaluates and compares the performance of three ML algorithms: Decision Trees (DT), Random Forests (RF), and Gradient Boosting Classifier (GBC), applied to analyze a dataset focused on fire type classification within the Kingdom of Bahrain. This study offers a novel approach for fire type classification, specifically tailored to the geographical conditions of Bahrain. The RF model achieved the highest performance, with accuracy (92.72%), recall (90.95%), precision (92.71%), F1 Score (91.74) and ROC-AUC (98.03%). This research establishes a foundation for utilizing advanced algorithms in fire analysis for Bahrain and demonstrates the effectiveness of RF in classifying the fire's type in similar contexts.