Figure 4

Communities detected in the mobile phone networks, at the disaggregated level of towers \(M_{0}\) (note that although towers are characterised by a single point, for the visual depiction we represent them by the Voronoi polygone associated to it), and the aggregated level of municipalities \(M_{m}\). The normalised mutual information, NMI, between partitions of network \(M_{0}\) and \(M_{m}\) is 0.64. Thus, the similarity between the communities found on the two levels of aggregation is higher than the similarity observed in the Twitter network between the disaggregated network of users (\(N_{0}\)), and the aggregated versions (see Fig. 5). The timescale parameter ρ is set to 0.75, as suggested by another study on the same dataset [27]