Authors: Pilla Rushitha, Puppala Pradeep, Yerrapatruni Jagadeesh Kumar
Abstract: Natural disasters such as floods, earthquakes, cyclones, droughts, landslides, and wildfires continue to pose significant threats to human life, infrastructure, and environmental sustainability. The growing complexity of climate patterns and environmental changes has increased the need for intelligent disaster prediction systems capable of providing accurate and timely forecasts. This project presents a Neural-XGBoost Hybrid Framework for Disaster Prediction and Management that integrates deep learning-based feature extraction with the robust classification capability of Extreme Gradient Boosting (XGBoost). The proposed approach utilizes disaster-related environmental and meteorological data, including rainfall, temperature, humidity, wind speed, and atmospheric conditions, to identify potential disaster events. Data preprocessing techniques such as cleaning, normalization, and feature selection are employed to enhance data quality and model performance. The neural network component automatically learns complex patterns and hidden relationships within the dataset, while XGBoost performs efficient multi-class disaster classification. Experimental evaluation demonstrates that the hybrid framework achieves superior prediction accuracy, improved generalization capability, and reduced overfitting when compared with conventional machine learning approaches. The system supports disaster preparedness, risk assessment, resource planning, and early warning mechanisms, enabling authorities to make informed decisions and minimize disaster-related losses. The proposed framework offers a scalable, reliable, and data-driven solution for modern disaster management applications.
DOI: http://doi.org/
