Understanding Twitter users' behaviour by social network analysis during disasters

Abstract

Fast response and assistance is a crucial step in disaster situations. With the use of the internet and social media in recent times, moving disaster management to digital platforms has become logical and interesting. Besides, the fact that social media plays a big role in daily usage for the majority of people provides many conveniences both during and after the disaster and even in the process of preparing for disaster. In the literature, it has been proved that social media plays an effective role in a natural disaster situation. However, no implementation has been made to use this interaction so far. Piece by piece there are many articles that are studied on disaster management. From these works, it is seen that the results of these implementation made are positive for disaster management ideas. Social media, especially Twitter, has a very big reaction in terms of disaster time. Both of the communities who are affected and non-affected from the disaster would like to share their thoughts, beliefs. People tend to share the news they have found about the disaster and sometimes these news may not trustable. Unfortunately, sometimes these shares might direct sufferers in the wrong way. In order to manage and correct these mistakes, it is necessary to analyze the reliability and reputation of users on social media. In this thesis, Social Network Analysis was applied in user "reply" interaction on Twitter. The main aim is to find a distinct user list and/or categories that can be useful in disaster management by sharing important messages. In disaster moments, accurate messages should be spread to the majority of the community in the fastest way. For this, we analyzed the behavior of social media before, during, and after the disaster moment and see if there is a significant user category to influence. We first retrieved raw data from Twitter and pre-processed the necessary attributes. We then created a node and edge tables which are essential for social network analysis. With the selected centrality measurements, we have compared and analyzed user categories and decide on important users. In the first part of the thesis, a literature review and prior works about the topic have been mentioned. Related works and their solutions were discussed. From the related works, the most similar three articles were mentioned and the significant difference in our study was stated. The purpose of the study and after implementation, what solution will be get was explained. All possible research questions were stated and the hypothesis was shown. After the literature part, the term Social Media Analytics was detailly explained with equations.