Dataset

GitHub Contributions

Open-source contribution data derived from the Google BigQuery GitHub public dataset, with synthetic regressors and outcome.

All datasets
Bipartite incidence pattern for GitHub Contributions

Graph Summary

Sparse cross-language/project structure makes this a difficult real benchmark.

The primary benchmark graph is built from unique (id1, id2) pairs. The figure shows a binned sparsity pattern of the bipartite incidence block, with both partitions relabeled to contiguous integer identifiers. This block is the off-diagonal part of the corresponding graph Laplacian.

548,843rows
41,751unique edges
27,970id1 levels
24,358id2 levels
13,232components
52,328nodes
CSV ZIP Stata DTA

Variables

Columns in the clean CSV:

id1 id2 id3 t len_commit_msg num_files x1 x2 y

The v1 graph uses id1 and id2. Additional identifier-like columns available for richer specifications: id3, t.

Source Notes

Google BigQuery public dataset `bigquery-public-data:github_repos`.

Google BigQuery public GitHub dataset documentation.

Historical Benchmark

2017 SEC benchmark timings for this dataset, in seconds:

MethodCitationSeconds
MAP-Aitken (Guimaraes 2012) 268.7
MAP-SD (Gaure 2013) 110.1
MAP-CG-Sym (Correia 2016) 114.4
MAP+Prune (Correia 2016) 127.1
LSMR (Gomez 2016) 169.6