With the spread of the Internet and the increase in access to the Internet via smartphones, the number of social networks to which users participate and the time they spend in social networks are increasing day by day. Each social network used by users for different purposes contains different user data. Finding users' accounts in different social networks and compiling the data found and compiling them into a single repository will be a very important factor that will both improve the operation of the recommended systems and increase the user experience.
In this thesis, it is aimed to collect the data of thousands of users in nine different social networks and to determine the accounts. Within the scope of the study, original node alignment and node similarity methods are proposed. While using the anchor method in topological-based node proposition, density relationships between the links are also taken into consideration. In the similarity-based node similarity method, attribute selection criteria, starting point detection problem, and variable formulation have increased the number of successful node matching. However, in this thesis, alignment and similarity were determined both according to the profile characteristics of the users and the relationships with other users.
Nine different methods have been proposed to find the same accounts on different social networks. It has been tested on two and six social networks, and match success rates of users have been measured. Success rates of up to 95% were achieved in these results. Thus, it is possible to create a full user profile covering multiple social networks for users whose different attributes are gathered on the same graph in multiple social networks. |