Jagrukta: SVM-Based Disaster Forecasting System

Abstract

This paper introduces Jagrukta, an intelligent disaster management and early warning system that leverages a Support Vector Machine (SVM) model trained on historical weather and geophysical data to predict and classify natural disasters such as floods, tsunamis, and earthquakes. Designed with the goal of enhancing disaster preparedness and minimizing the impact of calamities, Jagrukta processes decades of meteorological and seismic records—including rainfall intensity, oceanic temperature variations, tide levels, atmospheric pressure shifts, and seismic wave patterns—sourced from national and international datasets. These features are preprocessed and used to train an SVM classifier capable of categorizing input conditions into normal, alert, or critical states. The model achieves an average accuracy of 89%, showing high precision in forecasting flood and tsunami events, and delivering promising results for earthquake detection through temporal seismic pattern recognition. The system includes a real-time visualization interface that maps predictions geographically, enabling early alerts and faster response by disaster management authorities. Unlike resource-intensive deep learning models, Jagrukta’s SVM-based approach is interpretable, efficient, and optimized for deployment in regions with limited computational infrastructure. This project demonstrates the practical application of machine learning in disaster mitigation and aims to empower communities with timely, reliable, and data-driven insights. Future developments include incorporating dynamic forecasting through deep learning models, expanding the regional dataset, supporting mobile-based alerts, and integrating multilingual support to reach broader populations across India.