Flood prediction is a challenging task due to the complex and interdependent spatial relationships between rain- fall patterns, terrain characteristics, and river flow dynamics. Traditional machine learning models typically treat geographical regions as independent entities, which limits their ability to capture how flood risk propagates across neighboring and hydrologically connected areas. To overcome this limitation, this paper presents an intelligent flood risk prediction system based on Knowledge Graphs (KG) and Graph Neural Networks (GNN). A Spatial Knowledge Graph is constructed for the state of Kerala, modeling districts as nodes and physical adjacency as edges. Historical flood warning data is fused with static topological features to form a 6-dimensional spatio-temporal dataset. A Graph Convolutional Network (GCN) is trained using mathematically balanced class weights to mitigate dataset imbalance. The model outputs categorical risk levels and computes a continuous Flood Risk Index (0-100). Experimental evaluations demonstrate an exact accuracy of 86.90% and a relaxed close-warning accuracy of 93.71%. Furthermore, the AI model is successfully deployed into a production-ready Django web architecture featuring live OpenWeather API integration, interactive GeoJSON mapping, and an automated email alert engine, providing a comprehensive framework for proactive disaster management.
Flood Prediction, Knowledge Graph, Graph Neural Networks, GCN, Disaster Management, Django, Spatio- Temporal Modeling.
IRE Journals:
P. M. Tanay Reddy, M. Laxmikanth Reddy, M. Praneeth Kumar Yadav, T. Prudhvi Nandan Reddy "Intelligent Prediction of Flood Disaster Risk Levels Based on Knowledge Graph and Graph Neural Networks" Iconic Research And Engineering Journals Volume 9 Issue 9 2026 Page 902-906 https://doi.org/10.64388/IREV9I9-1714854
IEEE:
P. M. Tanay Reddy, M. Laxmikanth Reddy, M. Praneeth Kumar Yadav, T. Prudhvi Nandan Reddy
"Intelligent Prediction of Flood Disaster Risk Levels Based on Knowledge Graph and Graph Neural Networks" Iconic Research And Engineering Journals, 9(9) https://doi.org/10.64388/IREV9I9-1714854