Intelligent Forecasting of Flooding Intensity Using Machine Learning

This innovative study addresses critical flood prediction needs in Bor County, South Sudan, utilizing machine learning to develop an intelligent forecasting model. The research integrates diverse analytical techniques, including land use analysis and rainfall calculations, with a decade of weather data to understand complex hydrological dynamics. This research employs machine learning classifiers such as Support Vector Machines, Decision Trees, and Neural Networks. Findings reveal promising results, with the Linear SVM classifier achieving 87.5% prediction accuracy for raw data and 100% accuracy for high-velocity flooding events. The Naive Bayes classifier matched this performance, while Artificial Neural Networks showed a slight advantage in runoff estimation. The study's novelty lies in its holistic approach, combining machine learning with advanced visualization tools and geographic information systems. This creates a dynamic, real-time forecasting system bridging sophisticated analysis and practical flood management strategies. Focusing on model interpretability and multi-scale forecasting enhances its value to policymakers and disaster management authorities. This research significantly advances the application of AI to flood prediction and disaster management in offering future studies on humanitarian challenges. By enhancing early warning capabilities, this system substantially reduces flood-related losses and transforms disaster preparedness in vulnerable regions worldwide, potentially saving lives and mitigating economic impacts. Doi: 10.28991/CEJ-2024-010-10-010 Full Text: PDF