AI-Based model for site-selecting earthquake emergency shelters

Abstract

The objective of this research is to develop an AI-based model that provides a map of potential emergency shelters for seismic-prone regions. While the predominant approach for locating emergency shelters involves expert-answered questionnaires, their accuracy has often been criticized globally. To address this, machine learning can enhance the speed and accuracy of shelter site selection; the results can also be generalized to other regions. Support vector machine (SVM), K-nearest neighbor (KNN), logistic regression (LR), gaussian processes classifier (GPC), and artificial neural network (ANN) methods are used to develop the model here. These algorithms are trained using maps of emergency shelters in San Francisco, enabling the resulting model to automatically provide potential emergency shelter maps not only for the studied case but also for cities with similar criteria. Except for LR, the other algorithms achieved F1 scores between 0.7 and 1.0 in selecting emergency shelter sites. The model developed in this research can serve as a reliable tool for disaster management planners in managing emergency shelters for people affected by earthquakes.