Intelligent Prediction of Flood Disaster Risk Levels Based on Knowledge Graph and Graph Neural Networks

Flash flood disasters pose a significant threat to human life and property, making accurate prediction of risk levels crucial for disaster prevention and mitigation. This study introduces an innovative artificial intelligence approach based on knowledge graphs and graph neural networks. The method integrates multi-source data to construct a knowledge graph, which is then modeled using graph neural networks. We evaluate the model's performance using metrics such as accuracy, precision, recall, F1 score, and AUC. Under five-fold cross-validation, the AUC reached 0.84, with all performance indicators showing good results, indicating significant performance improvement. Experimental results demonstrate high prediction accuracy when tested on a dataset containing 9000 records. Compared with the three classical models in traditional machine learning, such as RF, SVM and ANN, the performance of this model is improved, and it is better than the traditional model. Through case analysis, risk levels in multiple regions were accurately predicted. Additionally, statistical analysis of flood disaster warning levels and flash flood risk zoning across cities in Jiangxi Province provides a visual representation of flood risk distribution and risk level proportions in different cities, offering strong reference for flood prevention, disaster mitigation, and urban planning. This method provides important scientific support for precise flash flood disaster prediction and risk management.