AI-Based model for site-selecting earthquake emergency shelters

Area of Science:

• Disaster Management
• Artificial Intelligence
• Geographic Information Systems

Background:

• Traditional emergency shelter location relies on expert questionnaires, which face global accuracy criticisms.
• Enhancing the speed and accuracy of shelter site selection is crucial for effective disaster response.
• AI and machine learning offer potential solutions for optimizing emergency shelter placement.

Purpose of the Study:

• To develop an AI-based model for mapping potential emergency shelters in seismic-prone regions.
• To improve the accuracy and efficiency of emergency shelter site selection compared to existing methods.
• To create a generalized model applicable to various regions with similar seismic criteria.

Main Methods:

• Utilized machine learning algorithms including Support Vector Machine (SVM), K-nearest neighbor (KNN), Logistic Regression (LR), Gaussian Processes Classifier (GPC), and Artificial Neural Network (ANN).
• Trained the models using existing emergency shelter maps from San Francisco.
• Evaluated algorithm performance based on F1 scores for accurate shelter site selection.

Main Results:

• Most algorithms (SVM, KNN, GPC, ANN) achieved high F1 scores between 0.7 and 1.0 in identifying potential emergency shelter sites.
• The developed AI model demonstrated the capability to automatically generate potential emergency shelter maps.
• The model's effectiveness was validated using San Francisco data, showing potential for generalization.

Conclusions:

• The AI-based model provides a reliable and accurate tool for disaster management planners in seismic-prone areas.
• Machine learning significantly enhances the speed and accuracy of emergency shelter site selection.
• The developed model can be generalized to assist in planning emergency shelters for earthquake-affected populations in similar urban environments.