Frequent subgraph discovery in dynamic networks

Bianca Wackersreuther, Peter Wackersreuther, Annahita Oswald, Christian Böhm, Karsten Borgwardt

2010 · 58 citations · 19 references

Concepts

TL;DR

Graphs model entities and relationships, and graph mining detects patterns, but most techniques target static graphs, whereas recent work increasingly addresses time‑varying graph sequences. The paper proposes a novel framework for frequent subgraph discovery in dynamic networks. The framework handles dynamic graphs with edge insertions and deletions over time and allows existing subgraph mining algorithms to be integrated for dynamic analysis. Extensive experiments on a large real‑world case study confirm the practical feasibility of the proposed framework.

Abstract

In many application domains, graphs are utilized to model entities and their relationships, and graph mining is important to detect patterns within these relationships. While the majority of recent data mining techniques deal with static graphs that do not change over time, recent years have witnessed the advent of an increasing number of time series of graphs. In this paper, we define a novel framework to perform frequent subgraph discovery in dynamic networks. In particular, we are considering dynamic graphs with edge insertions and edge deletions over time. Existing subgraph mining algorithms can be easily integrated into our framework to make them handle dynamic graphs. Finally, an extensive experimental evaluation on a large real-world case study confirms the practical feasibility of our approach.

References

19