Transition of Local Structures of Friendship Networks

Systematically collected class attendance records of students in university let us possible to observe friendship networks among students through time. With the friendship networks observed by this method we study local structures, which are called ego-centric networks, of friendship networks and transition of the structures. In order to observe transition of egocentric networks we classify them to clusters. For classification we examine a top-down cluster model, in which we define clusters based on social characters, and a hierarchical clustering method using structural attributes for ego-centric networks. The two methods are applied to real friendship networks that we observed for seven years and we see some observation about friendship transition. From our experiments the top-down model gives understandable transition of friendship structure. We can observe the process of growing friendship of students. The observation will be a promising approach to a computational analysis of construction of social roles in community.