Due to the growing popularity of social networking platforms, the analysis of online communities became in recent years a popular topic in different research fields. However, an aspect that has received only little attention is the question how the temporal aspects of social networks can be studied. This thesis bridges the gap and deals with the analysis of community structures in large social networks and their temporal dynamics. Two clustering techniques are proposed to detect communities in social networks and to study the evolution of these structures over time. The two approaches basically differ in the underlying definition of what constitutes a community over time: In the first case, a community is considered a subgroup that can be observed over time and a hierarchical edge betweenness clustering approach is proposed to detect such communities. In the second case, a community is defined as a subgroup that evolves over time and an incremental density-based clustering algorithm is proposed to detect and track these evolving communities. The applicability of the proposed approaches is evaluated by applying the methods to different real world data sets. The obtained results indicate that the introduced approaches are appropriate to efficiently detect online communities in large social networks and to track their evolution over time.
social networks, community analysis, network dynamics, data mining, graph clustering, online communities