Mining Frequent Sequential Subgraph Evolutions in Dynamic Attributed Graphs

Abstract

Mining patterns in a dynamic attributed graph has received more and more attention recently. However, it is a complex task because both graph topology and attributes values of each vertex can change over time. In this work, we focus on the discovery of frequent sequential subgraph evolutions (FSSE) in such a graph. These FSSE patterns occur both spatially and temporally, representing frequent evolutions of attribute values for general sets of connected vertices. A novel algorithm, named FSSEMiner, is proposed to mine FSSE patterns. This algorithm is based on a new strategy (graph addition) to guarantee mining efficiency. Experiments performed on benchmark and real-world datasets show the interest of our approach and its scalability.