Community detection in temporal networks using multi-layer technique
In this method the temporal graph is divied into layers according to their time stamps. The timeline is divided into time steps that form a network layer of a period of time.
Since a node can be a member of communities of other time-step layers than the time-step which is assigned to it, to solve this a edge's existance in one layer is partially transfered to the layer of previous time-step (if it exists) and to the layer of next time-step (if it exists).
Running the code:
Creating e-graph for temporal network:
python getGraph.py dataset_name
Creating e-graph for temporal network:
python getGraph.py dataset_name
Getting communities from e-graph:
python run.py dataset_name
Evalutation of communities:
python evaluation.py dataset_name
After partial transfer of edges between layers multi-layer community detection is applied and properties of detected communities are measured, results provided in the table below.
| Sno. | Datasets | Clustering Coeff. | Unifiability | Isoability |
|---|---|---|---|---|
| 1. | Eu-Core | 0.322403 | 0.005003 | 0.035551 |
| 2. | Eu-Core-D1 | 0.452008 | 0.012224 | 0.112353 |
| 3. | Eu-Core-D2 | 0.623362 | 0.038455 | 0.163557 |
| 4. | Eu-Core-D3 | 0.450960 | 0.073655 | 0.100457 |
| 5. | Eu-Core-D4 | 0.638279 | 0.035874 | 0.171590 |