Disaster Management Redefined: Integrating SVM-AE Techniques with Remote Sensing and Meteorological Data

This study presents a novel hybrid model, the Support Vector Autoencoder (SVAE), designed to enhance disaster management through the integration of Support Vector Machines (SVM) and Autoencoders (AE). By leveraging the strengths of both machine learning techniques, the SVAE model offers improved accuracy and reliability in predicting and managing natural disasters. The methodology involves comprehensive data collection from Sentinel-2 satellite imagery and Global Precipitation Measurement (GPM) mission data, supplemented by historical disaster records from the Emergency Events Database (EM-DAT). After rigorous preprocessing, key features such as the Normalized Difference Vegetation Index (NDVI), land surface temperature (LST), soil moisture content, and various meteorological parameters are extracted. These features are then normalized and used to train the SVM for supervised learning and the AE for unsupervised learning. The outputs of these modules are integrated through a fusion layer, which combines classification scores and anomaly detection signals to generate a final risk score. Performance comparison with other models, including Random Forest, k-NN, Decision Tree, and Naive Bayes, demonstrates that the SVAE model achieves superior accuracy, precision, recall, and F1-score. The proposed model’s accuracy reaches 97%, significantly outperforming other techniques in anomaly detection and risk assessment. The results indicate that the SVAE model is a robust tool for enhancing disaster preparedness and mitigation efforts, providing timely and actionable insights to decision-makers.