User-User Relationship Migration Observed in Communication Activity

Many Twitter users build various relationships through communication activity such as replies and retweets. For example, friends engage in conversations through replies. Fans unilaterally send many replies to celebrities. In this paper, we focus on such relationships between users. We assume that such relationships are classified into several patterns based on the feature values of communication reciprocity, and the relationships migrate to other ones as time progresses. We clarify the major relationships and transitions by analyzing the pattern frequency and transitions with high probability. From analysis results using a large amount of user pairs that we obtained over a long period, we detected several major and calm relationships.