The graph structure in dynamic networks changes rapidly. By leveraging temporal information inherent in network connections, models of these dynamic networks can be constructed to analyze how their structure changes over time. However, most existing generative models for temporal graphs grow networks, i.e. links are added over time, ignoring that real networks are dynamic with frequent lulls in activity. To this end, motifs have been established as building blocks for the structure of networks, thus modeling these higher-order structures can help to generate the graph structure seen on real-world networks. Furthermore, motifs can capture correlations in node connections and activity. To date, there are few dynamic-graph generative models and a minority of these consider higher-order network structure (instead of only node pair-wise connections). Such models have been evaluated using static graph structure metrics without incorporating measures that reflect the temporal behavior of the network. Our proposed DYnamic MOtif-NoDes (DYMOND) model considers both the dynamic changes in overall graph structure using temporal motif activity and the roles nodes play in motifs (e.g., one node plays the hub role in a wedge, while the remaining two act as spokes). We compare our model against three dynamic graph generative model baselines on real- world networks. We also propose a new methodology to adapt graph structure metrics to include the temporal aspect of the network. Our contributions in this paper are: (1) a statistical dynamic-graph generative model that samples graphs with realistic structure and temporal node behavior using motifs, and (2) a novel methodology for comparing dynamic- graph generative models and measuring how well they capture the underlying graph structure distribution and temporal node behavior of a real graph.

The Web Conference is announcing latest news and developments biweekly or on a monthly basis. We respect The General Data Protection Regulation 2016/679.