Efficient Flood Prediction with SVM and RF Algorithm

Abstract

Flood is a high risk of natural disasters such as floods due to its geological location at the intersection of four major tectonic plates. This study aims to predict flood risks using the Support Vector Machine (SVM) and Random Forest (RF) algorithms, utilizing rainfall, topography, and land use data. Historical rainfall data were obtained from BMKG, topographic data from GIS, and land use data from satellite imagery. The evaluation results show that the RF algorithm outperforms SVM, achieving 92.1% accuracy and an F1-score of 91.8%. RF has proven effective in capturing non-linear relationships between features influencing flood risk. This predictive system is expected to aid disaster mitigation, spatial planning, and the development of an early flood warning system.