Flood and Landslide Prediction and Risk Assessment System using Deep Learning with GSM-based Real Time Alerts

Landslides, floods, and other natural disasters cause severe damage to property, infrastructure, and human lives. Early prediction and real-time updates can help mitigate these impacts. This project introduces the Global Disaster Aggregation and Flood & Landslip Categorization System, a machine learning-based disaster management solution that provides real-time alerts and updates. The system features a Flask-built web application for monitoring live disaster warnings and integrates live news aggregation from reliable sources for timely updates. A key functionality is its GSM-enabled communication module, which sends real-time SMS notifications to high-risk locations, enabling individuals and organizations to take proactive safety measures. Additionally, an audio alarm mechanism emits a distinctive beep upon detecting threats, ensuring immediate awareness. The system employs Convolutional Neural Networks (CNNs) trained on a large dataset of ground-level images to classify disaster risks accurately. The model analyses soil displacement, water overflow patterns, and terrain instability, improving disaster preparedness. By integrating real-time alerts, beep notifications, and GSM-based SMS dissemination, the system ensures effective early warnings and emergency response. The CNN model achieved 96.8% training accuracy and 92.7% test accuracy, surpassing traditional statistical models. The real-time GSM-based alert system delivered notifications within 2.3 seconds, while audio alerts responded instantly. Comparative analysis confirmed that deep learning-based classification enhances disaster risk assessment and response efficiency. This research presents a scalable and effective solution for early disaster detection and risk mitigation. Future improvements include satellite imagery integration, IoT sensor data fusion, and advanced predictive analytics to enhance disaster preparedness and response strategies.