Author : Godavari Modalavalasa

Natural catastrophes pose a serious threat to property, human life, and vital infrastructure wherever they are found. The need to create effective disaster management systems derives from the growing frequency and intensity of disasters. This study explores the application of machine learning (ML) and deep learning (DL) techniques for disaster detection and classification in order to enhance disaster preparation and response. In this study, a comprehensive dataset that combines satellite images, meteorological data, and historical catastrophe records is used to investigate predicting natural disasters using Convolutional Neural Networks (CNN). The CNN model performs quite well, attaining 97.27% accuracy, 97.79% precision, 98.15% recall, and 97.97% F1-score. With accuracies of 95.33% and 95.23%, respectively, these results greatly outperform those of conventional models like Logistic Regression (LR) and Vision Transformer (ViT-B-32). A detailed evaluation, including loss and accuracy graphs, confirms the model’s efficient learning and stable convergence. These findings highlight CNN’s potential as a superior approach for natural disaster prediction, offering improved precision and dependability for disaster preparedness and early warning systems.

Keywords: Natural disaster, Disaster Risk Management, Disaster Prediction, Disaster Recovery, Machine Learning, Natural Disaster Dataset, Satellite Imagery, Meteorological Data, Disaster Management