Table 1. Comparison of LinkPred against the most important free/open-source link prediction software packages.
| Functionality | LinkPred | NetworkX | linkprediction | GEM | SNAP | linkpred | scikit-network |
|---|---|---|---|---|---|---|---|
| Supported languages | C++, Python (a subset of the functionalities), Java (a subset of the functionalities) | Python | R | Python | C++, Python (a subset of the functionalities) | Python | Python |
| Topological similarity methods | Yes (with shared memory and distributed parallelism) | Yes (no parallelism) | Yes (no parallelism) | No | No (A limited number of algorithms is included as an experimental component) | Yes (no parallelism) | Yes (no parallelism) |
| Global link prediction methods | Yes (with shared memory parallelism and for some predictors also distributed parallelism) | No | No | No | No | Yes (Rooted PageRank, SimRank, Katz, shortest path) | No |
| Graph embedding algorithms | LLE, Laplacian Eigenmaps, Graph Factorization, DeepWalk, LINE, LargeVis, node2vec, and HMSM | No | No | LLE, Laplacian Eigenmaps, Graph Factorization, HOPE, SDNE, and node2vec | node2vec and GraphWave | No | Spectral, SVD, GSVD, PCA, Random Projection, Louvain, Hierarchical Louvain, Force Atlas, and Spring. |
| Classifiers | Yes (mainly via mlpack) | No | No | No | No | No | No |
| Similarity measures | Yes | No | No | Yes | Yes | No | No |
| Test data generation | Yes | No | No | No | No | No | No |
| Performance measures | Yes | No | No | No | No | No | No |