A Unified Neural Network and XGBoost Learning Model for Disaster Prediction and Management

Abstract

Accurate disaster prediction plays a critical role in disaster management and mitigation by reducing the loss of life and property. Traditional machine learning approaches often struggle to handle complex disaster patterns, class imbalance in datasets, and challenges in multi-class disaster classification, which reduces prediction reliability. To address these issues, a hybrid disaster prediction framework based on Neural Networks and XGBoost is proposed for the multi-class classification of natural disasters. In the proposed framework, neural networks are employed for high-level feature extraction from real-world disaster data, while XGBoost (Extreme Gradient Boosting) performs efficient classification using gradient-boosted decision trees. The dataset, sourced from the EM-DAT International Disaster Database, comprises 4,543 samples with 31 features encompassing wildfires, floods, and earthquakes. The Synthetic Minority Oversampling Technique (SMOTE) is applied exclusively to the training data to address class imbalance and improve the model\'s generalization capability. Experimental results demonstrate improved performance, with the proposed Neural-XGBoost (N-XGB) model achieving 94.72% accuracy, 98.57% ROC-AUC, 90.49% F1-score, and 89.65% precision, outperforming traditional machine learning models. The proposed system provides a reliable and efficient disaster prediction approach that supports better disaster preparedness and decision-making.

Introduction

This study presents a hybrid machine learning model, called Neural-XGBoost (N-XGB), for predicting natural disasters such as floods, wildfires, and earthquakes. As natural disasters become more frequent and severe, accurate prediction is essential for disaster preparedness, resource allocation, and risk reduction. Traditional prediction methods often rely on historical analysis and expert judgement, making them less effective when handling large and complex datasets. To overcome these limitations, the proposed model combines the feature extraction capabilities of Neural Networks with the classification strength of XGBoost.

The model is developed using data from the EM-DAT International Disaster Database, which contains global records of natural disasters. The dataset includes 4,543 samples with 31 features, such as disaster type, location, magnitude, economic damage, and socio-economic factors. Data preprocessing involves handling missing values using median and mode imputation, label encoding categorical variables, and normalising numerical features with StandardScaler. To address class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) is applied to the training data, ensuring better representation of minority disaster classes.

The literature review highlights previous research showing that hybrid AI models, ensemble learning, and oversampling techniques improve disaster prediction accuracy. Studies on XGBoost demonstrate its efficiency in handling complex tabular data, while SMOTE has proven effective in improving minority-class classification. These findings support the development of a unified hybrid model that combines neural feature extraction, gradient boosting, and class balancing.

The proposed methodology consists of several stages: data collection, preprocessing, neural network-based feature extraction, SMOTE-based class balancing, XGBoost classification, and model evaluation. The neural network automatically learns high-level, nonlinear features from the disaster data, reducing the need for manual feature engineering. These extracted features are then classified by XGBoost, which builds multiple decision trees to produce accurate multi-class predictions. The implementation also includes a Streamlit-based web application that allows users to enter disaster-related data and receive real-time predictions.

Experimental evaluation shows that the proposed Neural-XGBoost (N-XGB) model significantly outperforms traditional machine learning algorithms, including Logistic Regression, Support Vector Machine (SVM), and Random Forest. The model achieved 94.72% accuracy, 90.49% F1-score, 89.65% precision, and a 98.57% ROC-AUC score, demonstrating excellent classification performance across all disaster types, including minority classes such as wildfires and earthquakes. The Streamlit application further confirms the model's practical usability by providing accurate single-sample and bulk predictions with high confidence. Overall, the proposed hybrid framework offers an effective and reliable solution for intelligent natural disaster prediction and decision support.

Conclusion

This paper presents a unified hybrid machine learning framework that combines Neural Networks and XGBoost for the multi-class classification and prediction of natural disasters including floods, wildfires, and earthquakes. The proposed Neural-XGBoost (N-XGB) model addresses three critical challenges in disaster prediction: complex nonlinear feature relationships, class imbalance in disaster datasets, and the limitations of standalone machine learning classifiers. By leveraging the feature extraction capability of deep neural networks and the ensemble classification power of XGBoost, combined with SMOTE-based data balancing, the system achieves a prediction accuracy of 94.72%, ROC-AUC of 98.57%, F1-score of 90.49%, and precision of 89.65%. The experimental results confirm that the proposed hybrid approach outperforms traditional machine learning models including Logistic Regression, SVM, and Random Forest, and demonstrates reliable classification performance across all disaster categories including minority classes. The deployment of the model through a Streamlit-based web interface further validates its practical applicability for real-time disaster prediction and decision support. The findings of this work provide a strong foundation for intelligent disaster prediction systems that can support preparedness planning, resource allocation, and emergency response across governmental, humanitarian, and environmental management domains.

References

[1] M. A. Saleem, W. Benjapolakul, W. Srisiri, S. Chaitusaney, and P. Kaewplung, \"A hybrid prediction model integrating artificial intelligence and geospatial analysis for disaster management,\" IEEE Access, vol. 13, pp. 43716–43727, 2025. [2] J. T. Thirukrishna, \"An enhanced discovery of multiple natural disasters using machine learning model,\" Earth Sci. Informat., vol. 18, no. 3, p. 324, Sep. 2025. [3] Universite Catholique de Louvain. (2024). CRED: Center of Research on the Epidemiology of Disasters. [Online]. Available: https://uclouvain.be/en/research-institutes/irss/cred.html [4] C. I. Donatti et al., \"Global hotspots of climate-related disasters,\" Int. J. Disaster Risk Reduction, vol. 108, Jun. 2024, Art. no. 104488. [5] E. Tung, A. Mostafavi, M. Li, S. Li, Z. Rasheed, and K. Shafique, \"A system for community lifeline and resource optimization in disaster management and planning,\" Int. J. Disaster Risk Reduction, vol. 119, Mar. 2025, Art. no. 105269. [6] G. Airlangga, \"Machine learning for tsunami prediction: A comparative analysis of ensemble and deep learning models,\" Kesatria: Jurnal Penerapan Sistem Informasi, vol. 6, no. 1, pp. 302–311, Jan. 2025. [7] S. Sharma and A. Gosain, \"Addressing class imbalance in remote sensing using deep learning approaches: A systematic literature review,\" Evol. Intell., vol. 18, no. 1, pp. 1–28, Feb. 2025. [8] N. V. Chawla, A. Lazarevic, L. Hall, and K. W. Bowyer, \"SMOTEBoost: Improving prediction of the minority class in boosting,\" in Proc. Eur. Conf. Princ. Data Mining Knowl. Discovery, 2003, pp. 107–119. [9] T. Chen and C. Guestrin, \"XGBoost: A scalable tree boosting system,\" in Proc. 22nd ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining, Aug. 2016, pp. 785–794. [10] R. L. Jones, A. Kharb, and S. Tubeuf, \"The untold story of missing data in disaster research: A systematic review of the empirical literature utilising the EM-DAT,\" Environ. Res. Lett., vol. 18, no. 10, Oct. 2023, Art. no. 103006. [11] P. Ghorpade et al., \"Flood forecasting using machine learning: A review,\" in Proc. 8th Int. Conf. Smart Comput. Commun. (ICSCC), Jul. 2021, pp. 32–36. [12] J. Jia and W. Ye, \"Deep learning for earthquake disaster assessment: Objects, data, models, stages, challenges, and opportunities,\" Remote Sens., vol. 15, no. 16, p. 4098, Aug. 2023. [13] G. Husain et al., \"SMOTE vs. SMOTEENN: A study on the performance of resampling algorithms for addressing class imbalance in regression models,\" Algorithms, vol. 18, no. 1, p. 37, Jan. 2025. [14] R. Cantini et al., \"Harnessing prompt-based large language models for disaster monitoring and automated reporting from social media feedback,\" Online Soc. Netw. Media, vol. 45, Jan. 2025, Art. no. 100295. [15] EM-DAT: The International Disaster Database. (2024). EM-DAT Data Repository. [Online]. Available: https://public.emdat.be/data

Copyright

Copyright © 2026 Ms. V. Bharathi , SK. Raqeeba. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.