Skip to main content

Python package for link prediction

Project description

linkpred is a Python package for link prediction: given a network, linkpred provides a number of heuristics (known as predictors) that assess the likelihood of potential links in a future snapshot of the network.

While some predictors are fairly straightforward (e.g., if two people have a large number of mutual friends, it seems likely that eventually they will meet and become friends), others are more involved.

linkpred can both be used as a command-line tool and as a Python library in your own code.


linkpred (v0.5 and later) works under Python 3.6, 3.7, and 3.8. Version 0.4.1 was the last to support versions 3.4 and 3.5. It depends on:

  • matplotlib

  • networkx

  • numpy

  • pyyaml

  • scipy

  • smokesignal

You should be able to install linkpred and its dependencies using pip (pip install linkpred or python -m pip install linkpred). If you do not yet have Python installed, I recommend starting with Anaconda, which includes optimized versions of packages like numpy. If you want to use the Community predictor, which relies on community structure of the network, make sure you also have the python-louvain package by installing with pip install linkpred[all].

Example usage as command-line tool

A good starting point is linkpred --help, which lists all the available options. To save the predictions of the CommonNeighbours predictor, for instance, run:

$ linkpred examples/ -p CommonNeighbours --output cache-predictions

where examples/ is a network file in Pajek format. Other supported formats include GML and GraphML. The full output looks like this:

$ linkpred examples/ -p CommonNeighbours --output cache-predictions
16:43:13 - INFO - Reading file 'examples/'...
16:43:13 - INFO - Successfully read file.
16:43:13 - INFO - Starting preprocessing...
16:43:13 - INFO - Removed 35 nodes (degree < 1)
16:43:13 - INFO - Finished preprocessing.
16:43:13 - INFO - Executing CommonNeighbours...
16:43:14 - INFO - Finished executing CommonNeighbours.
16:43:14 - INFO - Prediction run finished

$ head examples/inf1990-2004-CommonNeighbours-predictions_2016-04-22_16.43.txt
"Ikogami, K"    "Ikegami, K"    5.0
"Durand, T"     "Abd El Kader, M"       5.0
"Sharma, L"     "Kumar, S"      4.0
"Paul, A"       "Durand, T"     4.0
"Paul, A"       "Dudognon, G"   4.0
"Paul, A"       "Abd El Kader, M"       4.0
"Karisiddippa, CR"      "Garg, KC"      4.0
"Wu, YS"        "Kretschmer, H" 3.0
"Veugelers, R"  "Deleus, F"     3.0
"Veugelers, R"  "Andries, P"    3.0

Example usage within Python

>>> import linkpred
>>> G = linkpred.read_network("examples/")
11:49:00 - INFO - Reading file 'examples/'...
11:49:00 - INFO - Successfully read file.
>>> len(G)   # number of nodes
>>> # We exclude edges already present, to predict only new links
>>> simrank = linkpred.predictors.SimRank(G, excluded=G.edges())
>>> simrank_results = simrank.predict(c=0.5)
>>> top =
>>> for authors, score in top.items():
...    print(authors, score)
Tomizawa, H - Fujigaki, Y 0.188686630053
Shirabe, M - Hayashi, T 0.143866427916
Garfield, E - Fuseler, EA 0.148097050146
Persson, O - Larsen, IM 0.138516589957
Vanleeuwen, TN - Noyons, ECM 0.185040358711

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

linkpred-0.5.1.tar.gz (23.6 kB view hashes)

Uploaded source

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page