Skip to Main content Skip to Navigation
Theses

Overlapping community detection in dynamic networks

Qinna Wang 1, 2
2 DNET - Dynamic Networks
Inria Grenoble - Rhône-Alpes, LIP - Laboratoire de l'Informatique du Parallélisme
Abstract : In complex networks, the notion of community structure refers to the presence of groups of nodes in a network. These groups are more densely connected internally than with the rest of the network. The presence of communities inside a network gives an insight on network structural properties. For example, in social networks, communities are based on common interests, location, hobbies.... Generally, a community structure is described by a partition of the network nodes, where each node belongs to a unique community. A more reasonable description seems to be overlapping community structure, where nodes are allowed to be shared by several communities. Moreover, when considering dynamic networks whose interactions between nodes evolve in time, it appears crucial to consider also the evolution of the intrinsic community structure. This thesis focus on mining dynamic community evolution and overlapping community detection. We have proposed two distinct methods for overlapping community detection. The first one named clique optimization and the second one called fuzzy detection. Our clique optimization aims to identify granular overlaps and it is a fine grain scale approach. Our fuzzy detection is at a coarser grain scale with the strategy of identifying modular overlaps. Their applications in synthetic and real networks indicate that both methods can be used for characterizing overlapping nodes but in distinct and complementary views. We also propose the definition of predecessor and successor in mining community evolution. Such definition describes the relationship between communities at different time steps. We use it to detect community evolution in dynamic networks and show how modular overlaps evolve over time. A visualization tool called lineage diagrams is used to show community evolution by connecting communities in relationship of predecessor and successor. Several cases are studied.
Document type :
Theses
Complete list of metadatas

Cited literature [135 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00701217
Contributor : Abes Star :  Contact
Submitted on : Thursday, May 24, 2012 - 6:24:12 PM
Last modification on : Wednesday, November 20, 2019 - 2:59:19 AM
Long-term archiving on: : Saturday, August 25, 2012 - 2:45:09 AM

File

WANG_Qinna_2012_These.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-00701217, version 1

Collections

Citation

Qinna Wang. Overlapping community detection in dynamic networks. Other [cs.OH]. Ecole normale supérieure de lyon - ENS LYON, 2012. English. ⟨NNT : 2012ENSL0713⟩. ⟨tel-00701217⟩

Share

Metrics

Record views

1325

Files downloads

1903