Modeling and analyzing of academic collaboration of universities as social network

Together with rapidly evolving technology and human needs, the problems faced by people are becoming increasingly complex. The solution of this complicated problem becomes one that an individual cannot solve with knowledge. For this reason, the collaboration of research and development institutions (R&D), especially universities, as well as researchers working in different institutions have gained importance. The work done in this way is called as academic collaboration networks. Academic collaboration networks are in the status of developing social networks and are being studied within the scope of social network analysis (SNA). As for the social networking, the most commonly used modeling tool is the graphs, so the collaboration networks too. As for social networks, the most commonly used modeling tools are the graphs. Modeling is the expressing of large data sets using easily understandable diagrams, shapes, texts and symbols using software technologies. Different data types are modeled using different topologies. Like many other fields that deal with the relationship between actors and these actors, the analysis of collaboration, is modeled as a graph and expressed mathematically. The aim of this thesis is to model the academic collaboration between the universities of Turkey using graph topology and to make analyzes through the produced models. The data used in modeling was drawn with an application, developed for this study, from the database of Web of Science (WOS), one of the largest academic databases. The data obtained are stored as a relational database at the author, institution, paper information in center. The records in the database were queried and the desired analyzes were extracted. In addition, with the spectral graph partitioning method has examined how Turkish universities show a clustering of collaboration. Due to the data losses seen in the clusters with spectral graph partitioning, a look-alike community detection was applied on a regional basis and an algorithm was proposed to perform a noise cleanup in community discovery.