Figure 8

Example of the clustering algorithm. (A) Initial situation. Each node is an ad; nodes are linked if they are duplicates. The 1–7–8 and 2–3 subgraphs are fully connected; the 4–5–6–9–10 subgraph is not. (B) Final situation. We split the cluster 4–5–6–9–10 by removing the edge with minimal duplicate probability, in this case 4–5. The cluster 4–6–9–10 is still not fully connected, but sufficiently connected. As a condition for connectivity, we choose \(M\geq \frac{5}{6}\frac{N(N-1)}{2}\), where M is the number of edges and N the number of nodes in the cluster