GIS-based flood vulnerability mapping in a tropical river basin using analytical hierarchy process (AHP) and machine learning approach

Floods have become a significant global concern in recent times due to their devastating impact to infrastructure and society. The key flood-conditioning factors such as rainfall, elevation, slope, drainage density, distance from the river, TWI, soil type, and LULC were assigned weightage using the AHP method. Several machine learning models were employed to examine the relationship between flood occurrences and susceptibility. The flood risk zones were classified into five categories: very high (4%), high (14%), moderate (52%), low (16%), and very low (14%). The AUC-ROC results demonstrated high accuracy across different models: AHP (0.825), Random Subspace (0.974), Bagging (0.985), Radial Basis Function (0.937), Artificial Neural Network (0.953), M5P (0.946), Additive Regression (0.960), Random Forest (0.932), and Support Vector Machine (0.882). The findings of this study will assist relevant authorities in formulating flood mitigation strategies and supporting Sustainable Development Goals (SDGs).