NetworkX-Temporal

NetworkX-Temporal extends the NetworkX library to dynamic graphs, i.e., temporal network data.

This package provides new TemporalGraph classes, which inherit NetworkX’s graph classes and implement additional functions to manipulate temporal data within. Among others, it provides ways to slice() a graph into snapshots and transform or convert it to other libraries and formats.

Install

The package supports Python 3.7+ and is readily available from PyPI:

$ pip install networkx-temporal

Additional support for plotting graphs may be optionally installed with:

$ pip install 'networkx-temporal[draw]'

Quick start

The following is a quick example of the package in action, covering its basic functionality. More examples are accessible via the sidebar and also available as a Jupyter notebook (open on Colab).

Build and slice temporal graph

Create a directed TemporalGraph object and slice() it into a number of snapshots:

>>> import networkx_temporal as tx
>>>
>>> TG = tx.temporal_graph() # TG = tx.TemporalMultiGraph()
>>>
>>> TG.add_edge("a", "b", time=0)
>>> TG.add_edge("c", "b", time=1)
>>> TG.add_edge("d", "c", time=2)
>>> TG.add_edge("d", "e", time=2)
>>> TG.add_edge("a", "c", time=2)
>>> TG.add_edge("f", "e", time=3)
>>> TG.add_edge("f", "a", time=3)
>>> TG.add_edge("f", "b", time=3)
>>>
>>> TG = TG.slice(attr="time")
>>>
>>> print(TG)

TemporalMultiGraph (t=4) with 6 nodes and 8 edges

The attr parameter optionally defines the attribute name to use for slicing the temporal graph, while the number of snapshots to be created may likewise be specified with the bins parameter:

>>> TG.slice(attr="time", bins=2)

TemporalMultiGraph (t=2) with 6 nodes and 8 edges

Note that the total number of nodes may vary, while the total number of edges is preserved.

Note

The slice() method by default creates a snapshot for unique time value in the temporal graph. It internally stores views of the original graph, so no data is copied unless otherwise specified.

Plot snapshots

We may visualize the resulting temporal graph using the draw() function:

>>> tx.draw(TG, layout="kamada_kawai", figsize=(8, 2))

Save and load data

The write_graph() and read_graph() functions accept compressed temporal graphs:

>>> TG = tx.read_graph("temporal-graph.graphml.zip")
>>> tx.write_graph(TG, "temporal-graph.graphml.zip")

Both functions support the same extension formats as in the installed NetworkX library version.

Convert and transform graphs

This package allows to transform a TemporalGraph between different graph representations:

Static graphs: flattened or aggregated version of the temporal graph.
Snapshot-based temporal graphs: a list of node- or edge-level snapshots.
Event-based temporal graphs: a sequence of edge-level events (interactions).
Unrolled temporal graphs: a single graph with time-stamped nodes and edges.

>>> G = TG.to_static()        # TG = tx.from_static(G)
>>> STG = TG.to_snapshots()   # TG = tx.from_snapshots(SG)
>>> ETG = TG.to_events()      # TG = tx.from_events(EG)
>>> UTG = TG.to_unrolled()    # TG = tx.from_unrolled(UG)

In addition, both static and temporal graphs may be converted to the following graph formats:

>>> tx.convert(G, "igraph")

<igraph.Graph at 0x7f048ef52c50>

Links

For more information on using this package, please refer to the following sections:

Package reference for details on the classes and functions implemented by each module.
Examples covering some of its main functionalities and most common use cases.