TURNING DATA INTO ACTIONABLE INTELLIGENCE FOR FLOOD PREDICTION AND ADVISORY SYSTEM

Floods remain among the most destructive natural disasters, inflicting severe losses on infrastructure, property, and human lives. The development of accurate and timely flood prediction and advisory systems is therefore critical for disaster risk reduction and resilience building. This study investigates the transformation of heterogeneous raw data into actionable intelligence to support flood prediction and advisory processes. Leveraging advanced data analytics and machine learning (ML) techniques, we propose a Flood Prediction and Advisory System (FPAS) model that integrates diverse data sources, with particular emphasis on datasets from the Nigerian Meteorological Agency (NiMet). The research adopts a hybrid methodology that combines Object-Oriented Analysis and Design Methodology (OOADM) with the Cross Industry Standard Process for Data Mining (CRISP-DM). Three ML algorithms—Support Vector Machine (SVM), Random Forest, and XGBoost—were applied to the NiMet dataset, with Random Forest and XGBoost demonstrating superior performance in terms of accuracy, precision, and overall predictive ranking. The findings highlight substantial improvements in forecast reliability, underscoring the potential of data-driven systems to enhance decision-making for policymakers and emergency response agencies. The paper concludes by discussing practical implications for flood risk management and avenues for future research.