Presenting a Predictive Model of the Physical Vulnerability of Neighborhoods against Earthquakes (Case Study: District 1 of Tehran Municipality)

The issue of the physical vulnerability of neighborhoods against earthquakes is the subject of this research. Crisis management and smart crisis management are generally considered in three stages: before, during, and after the crisis. The management of a smart crisis in all three stages, with emphasis on preparedness and anticipation of vulnerability against disasters such as earthquakes, can provide a way to predict vulnerability and increase power in decision-making. The purpose of this research is to present a predictive model for the vulnerability of the physical context against earthquakes in District 1 of Tehran municipality using machine learning. The research method is analytical and quantitative. Some of the data was collected from GIS and some were extracted from the map analysis. According to the use of a supervised learning approach in this research, labeling was performed by researchers in five different spectrums. Decision tree algorithm, support vector machine (SVM), and multilayer neural network (MLP) were used as machine learning algorithms. The portion of training data for the test was considered to be 70 to 30. By examining the accuracy of the model by the confusion matrix, it was found that the decision tree algorithm with an accuracy of 99.50, sensitivity of 99.42, and error of 0.5 has better performance than the other two algorithms. Moreover, the neural network with an accuracy of 97.85, sensitivity of 97.57, and error of 2.15 showed better performance than support vector machine.