Real-Time Earthquake Detection and Risk Classification using Machine Learning and Streamlit

Earthquakes pose a significant threat to human life and infrastructure worldwide, with approximately 500,000 detectable events occurring each year. Early detection and rapid risk assessment are critical for disaster mitigation and emergency response. Recent advances in machine learning (ML) have shown great promise in automating earthquake detection. However, most existing systems are not deployed as real-time web applications accessible to the public. In this paper, we present a real-time earthquake detection and risk classification system that leverages live data from the United States Geological Survey (USGS) API. Our system employs a Random Forest classifier trained on four key features: magnitude, depth, latitude, and longitude. The model classifies each earthquake into three risk categories — HIGH, MEDIUM, and LOW — with an associated confidence score and a risk score on a 0-100 scale. The system integrates live data streaming, interactive visualizations including world maps, magnitude distribution charts, and seismic waveform signals. Additionally, the application features country-wise filtering, depth and time range filters, satellite imagery, and an admin panel for data management. Experimental results demonstrate 100% classification accuracy on the test dataset, with magnitude identified as the most important feature. The system is publicly deployed as a web application, making it accessible for disaster monitoring and research applications. Future work includes integration of deep learning models for improved prediction accuracy and real-time alert broadcasting.