Application of Multilayer Perceptron (MLP) Neural Network Model in Urban Vulnerability Zoning with Emphasis on Earthquake (A Case Study on Municipal District 20 in Tehran)

District 20 (Shahr-e-Ray), as the southernmost urban area among the municipal districts of Tehran, has a population of 4553740 individuals and an area of 22km2 within the urban zone and 178km2 outside this zone. The earthquake risk is estimated to be very high in this region due to the tectonic and geographical position, presence of numerous faults around the region, the occurrence of several historical destructive earthquakes within this area, as well as other tectonic and geological evidence. In the present study, by investigating the current status, analyzing and classifying the vulnerability of the habitats in this region, and using the MLP model, a new strategy is presented. The results of the model based on the input models indicated the higher accuracy and efficiency of the standard classification method compared to the standard max/min method. By taking a look at the map of the standard classification method in the applied model, it can be found that the orange and nearly blue spots, which are mostly scattered in the central part of the region, have the highest correlation with the worn texture and the highest vulnerability. According to the results of the model in terms of vulnerability expansion and zone, out of the total area of the region, 21% has the high and very high vulnerability, 61% medium vulnerability, and 18% low vulnerability. Also, the results of the population layers indicated high, medium, and low vulnerability intensities for 56.8%, 27.9%, and 14.1% of the total population, representing the population density in worn-out buildings with narrow pathways. Moreover, the results concerning the layers of material type and important places showed that the material types including metals, bricks, semi-metallic, …, respectively, and the important places such as educational centers, clinics, and medical centers have the highest vulnerability intensities