Cohesive Co-evolution Patterns in Dynamic Attributed Graphs Elise Desmier 1 , Marc Plantevit 2 , C´ eline Robardet 1 , and Jean-Franc ¸ois Boulicaut 1 1 Universit´ e de Lyon, CNRS, INSA-Lyon, LIRIS, UMR5205, F-69621, France 2 Universit´ e de Lyon, CNRS, Universit´ e Lyon 1, LIRIS, UMR5205, F-69622, France Abstract. We focus on the discovery of interesting patterns in dynamic attributed graphs. To this end, we define the novel problem of mining cohesive co-evolution patterns. Briefly speaking, cohesive co-evolution patterns are tri-sets of vertices, timestamps, and signed attributes that describe the local co-evolutions of similar vertices at several timestamps according to set of signed attributes that express attributes trends. We design the first algorithm to mine the complete set of cohe- sive co-evolution patterns in a dynamic graph. Some experiments performed on both synthetic and real-world datasets demonstrate that our algorithm enables to discover relevant patterns in a feasible time. 1 Introduction Real-world phenomena are often depicted by graphs where vertices represent entities and edges represent their relationships or interactions. With the rapid development of social media, sensor technologies and bioinformatics assay tools, large heterogeneous information networks have become available and deserve new knowledge discovery methods. As a result, graph mining has become an extremely active research domain. It has recently been extended into several complementary directions as multidimensional graphs [5], attributed graphs [16,17,22], and dynamic graphs [6]. Indeed, entities can be described by one or more attributes that constitute the attribute vectors associated with the graph vertices. Moreover, in many applications, edges may appear or disappear through time giving rise to dynamic graphs. So far, sophisticated methods have been designed to provide new insights from at- tributed or dynamic graphs. Recent contributions have shown that using additional in- formation associated to vertices enables to exploit both the graph structure and local vertex attributes [16,17,22]. Dynamic graphs have been studied in two different ways. On one hand, it is possible to study the evolution of specific properties (e.g., the di- ameter). On the other hand, it makes sense to look at local patterns and it provides a large spectra of approaches to characterize the evolution of graphs with association rules [3,18], transformation rules [24] or other types of patterns [6,12,13,15,20]. Analysing dynamic attributed graphs (i.e., sequence over time of attributed graphs whose relations between vertices and attributes values depends on the timestamp) has been less studied and we claim that it is interesting for several reasons. First, this kind of data offers a richer representation of real-world phenomena in which entities have their J.-G. Ganascia, P. Lenca, and J.-M. Petit (Eds.): DS 2012, LNAI 7569, pp. 110–124, 2012. c  Springer-Verlag Berlin Heidelberg 2012