C-Blondel: An Efficient Louvain-Based Dynamic Community Detection Algorithm

Evolutionary clustering and community detection algorithms for social media health surveillance

“…Some recent advances in evolutionary clustering techniques propose new algorithms for dynamic modularity-based community detection, like one utilized for the Louvain method in Section 4 . The C-Blondel algorithm recently introduced by Seifikar, Farzi, and Barati (2020) is one such method, being based on detecting communities in dynamic networks by building compressed graphs of the network and integrating the Louvain method to discover communities. Rather than the typical association of a snapshot cost SC and a temporal cost TC , with varying representation, as well as the trade-off parameter , the evolutionary aspect is satisfied by building-in the communities at : C ( ) and incorporating them into a compressed graph based on previous communities as well as network changes between previous and current snapshots ( Seifikar et al, 2020 ).…”

“…The C-Blondel algorithm recently introduced by Seifikar, Farzi, and Barati (2020) is one such method, being based on detecting communities in dynamic networks by building compressed graphs of the network and integrating the Louvain method to discover communities. Rather than the typical association of a snapshot cost SC and a temporal cost TC , with varying representation, as well as the trade-off parameter , the evolutionary aspect is satisfied by building-in the communities at : C ( ) and incorporating them into a compressed graph based on previous communities as well as network changes between previous and current snapshots ( Seifikar et al, 2020 ). The result of this compression includes nodes and edges of the history of the network incorporated into a new graph as super nodes and super edges ( Seifikar et al, 2020 ).…”

“…Rather than the typical association of a snapshot cost SC and a temporal cost TC , with varying representation, as well as the trade-off parameter , the evolutionary aspect is satisfied by building-in the communities at : C ( ) and incorporating them into a compressed graph based on previous communities as well as network changes between previous and current snapshots ( Seifikar et al, 2020 ). The result of this compression includes nodes and edges of the history of the network incorporated into a new graph as super nodes and super edges ( Seifikar et al, 2020 ). They additionally seek to solve the issue that temporal smoothing has been considered for when subtle node changes between generations end up destroying the community structure of the network.…”

“…They additionally seek to solve the issue that temporal smoothing has been considered for when subtle node changes between generations end up destroying the community structure of the network. This is accomplished by way of evaluating network changes that include appearing and disappearing nodes and edges from the network ( Seifikar et al, 2020 ). Degree centrality is heuristically used to determine what nodes could potentially destabilize the network based upon the product of a destruction parameter with average community degree ( Seifikar et al, 2020 ).…”

“…This is accomplished by way of evaluating network changes that include appearing and disappearing nodes and edges from the network ( Seifikar et al, 2020 ). Degree centrality is heuristically used to determine what nodes could potentially destabilize the network based upon the product of a destruction parameter with average community degree ( Seifikar et al, 2020 ). Bringing this all together, the destruction parameter fulfils the purpose of trading off modularity for execution time, and was found reduce the time complexity of the algorithm while sustaining comparable modularity to other methods S-Blondel and D-Blondel across three bench-marking datasets ( Seifikar et al, 2020 ).…”