Machine Learning for Earthquake Emergency Evacuation: Intelligent Site Selection and Neighborhood Navigation System

Abstract

Natural disasters such as earthquakes pose significant threats to human life and infrastructure, requiring efficient evacuation strategies to minimize casualties. Traditional evacuation systems rely heavily on static planning and manual decision-making, which often fail to adapt to dynamic disaster conditions. This research proposes a machine learning-based approach for earthquake emergency evacuation, focusing on intelligent site selection and neighborhood navigation.The system utilizes geographic data, including latitude and longitude, combined with historical disaster datasets to train multiple machine learning models such as Random Forest, Decision Tree, K-Nearest Neighbors (KNN), and Support Vector Machine (SVM). These models are designed to classify disaster-prone areas and predict suitable evacuation zones. By analyzing spatial patterns and disaster types, the system assists in identifying safe locations for evacuation.A graphical user interface (GUI) is developed using Python’s Tkinter library to facilitate user interaction. The interface allows users to upload datasets, train models, select preferred algorithms, and input geographic coordinates for real-time predictions. The system outputs predicted disaster types and suggests evacuation strategies accordingly.Additionally, the integration of geospatial visualization using Folium enhances the system by providing an interactive disaster map. Different disaster types are represented using distinct colors, enabling users to visually interpret risk zones and safe areas. This feature significantly improves decision-making during emergency scenarios.The proposed system aims to overcome limitations of traditional evacuation planning by incorporating real-time predictions and intelligent data analysis. It enhances accuracy, reduces response time, and supports authorities in making informed decisions. Experimental results indicate that ensemble models such as Random Forest achieve higher accuracy compared to other classifiers, making them suitable for deployment in disaster management systems. Overall, this research contributes to the development of smart disaster response systems by combining machine learning and geospatial technologies. The system can be extended further by integrating real-time sensor data, IoT devices, and mobile-based navigation to provide dynamic evacuation routes. This approach has the potential to save lives and improve resilience in disaster-prone regions.