Published January 1, 2026 | Version v1

A Hybrid Neural Network And Xgboost Framework For Accurate Disaster Prediction And Management

Description

Natural disasters such as floods, wildfires, and earthquakes cause severe damage to human life, infrastructure, and the global economy. As the frequency and intensity of these disasters continue to increase due to climate change and environmental factors, the need for reliable disaster prediction systems has become more critical than ever. Traditional disaster prediction methods often rely on historical patterns and statistical models, which are limited in their ability to handle complex and imbalanced datasets.This study proposes a hybrid machine learning framework that integrates Neural Networks and XGBoost to improve the accuracy and reliability of disaster prediction. In the proposed approach, neural networks are used to automatically extract meaningful and high-level features from disaster datasets, while XGBoost performs the final classification of disaster types. To address the common issue of class imbalance in disaster datasets, the Synthetic Minority Over-sampling Technique (SMOTE) is applied during data preprocessing.The model is evaluated using a real-world disaster dataset containing records of floods, wildfires, and earthquakes. Experimental results demonstrate that the proposed hybrid model significantly outperforms traditional machine learning techniques such as Random Forest, Support Vector Machines, and Logistic Regression. The proposed system achieves high prediction accuracy and improved F1-scores across all disaster categories.Overall, the hybrid Neural-XGBoost framework provides a robust and efficient solution for disaster prediction and management. The system can support disaster management agencies by enabling early warning systems, improving preparedness strategies, and assisting in better resource allocation during emergency situations.

Files

IJSET_V14_issue2_270.pdf

Files (562.5 kB)

Name Size
md5:dcb506e8bc2f9200777d1556df4dc8da
562.5 kB Preview Download

Additional details