Utilization of artificial neural networks for calculation of social vulnerability in disaster management

In recent years, artifical intelligence, which tries to mimic human thinking and decision making process in computerized environments, has been applied in different areas. New developments in technology and software areas allowed to broaden these different applications. In this thesis, the goal is to research artificial intelligence and it?s techniques and to apply them for calculation of social vulnerability in eartquakes. Especially Neural Networks and Fuzzy Logic are main focus areas.Earthquake damage, one of the common types of natural disasters in our country, can be scrutinized in three categories: physical, economic and social. Among these three physical vulnerability has been studied widely. However when it comes to social vulnerability, it is a new area and there has been limited number of studies conducted in this area. Rapid population growth, internal and external migration, factors such as lack of information and training constitute social vulnerability factors and if they are not improved over time, the risks of natural disaster increases. Therefore, social vulnerability is also important and it must be also researched in detail to avoid consequences.In order to reach this goal, first cities in Turkey are classified into groups based on social factors by using neural networks. Then, the relationship between created groups and social factors is established by using neural networks.For this purpose, first a comprehensive literature review about Artificial Intelligence, Artificial Neural Networks and Fuzzy Logic was conducted. Then, in order to classify cities into social groups SPSS Clementine was used. Finally artificial neural networks module of MATLAB has been used to complete the analysis. The data for social indicators of 81 cities were gathered from Turkish Statistics Institute.The result of this study provides a support that social factors can be used in order to meausure social vulnerability level of cities and Artificial Neural Networks can be utilized for this purpose.