Dynamic Local Community Detection with Anchors

Community detection is a challenging research problem, especially in dynamic networks, since in this case communities cannot remain stable as they evolve. In evolving networks new communities may emerge and existing communities may disappear, grow or shrink. There are many cases where someone is more interested in the evolution of a particular community, to which an important node belongs, rather than in the global partitioning of a dynamic network. However, due to the drifting problem where one community can evolve into a completely different one, it is difficult to track the evolution of communities. Our aim is to identify the community that contains a node of particular importance, called anchor, and its evolution over time. The framework we propose circumvents the identity problem by allowing the anchor to define the core of the relevant community partially or fully. Preliminary experiments with synthetic datasets demonstrate the positive aspects of the proposed framework in identifying communities with high accuracy.