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.
The latest release version of Linkpred (v0.5.1) works under Python 3.6, 3.7, and 3.8. Current master branch supports Python versions 3.8 to 3.11.
Linkpred 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].
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/inf1990-2004.net -p CommonNeighbours --output cache-predictions
where examples/inf1990-2004.net is a network file in Pajek format. Other supported formats include GML and GraphML. The full output looks like this:
$ linkpred examples/inf1990-2004.net -p CommonNeighbours --output cache-predictions
16:43:13 - INFO - Reading file 'examples/inf1990-2004.net'...
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>>> import linkpred
>>> G = linkpred.read_network("examples/training.net")
11:49:00 - INFO - Reading file 'examples/training.net'...
11:49:00 - INFO - Successfully read file.
>>> len(G) # number of nodes
632
>>> # 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 = simrank_results.top(5)
>>> 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