Local Pluralistic Homophily — Preprocessed Datasets

Authors: Fernando Barraza; Carlos Ramírez-Ovalle; Abel Álvarez; Alejandro Fernández
Contact: fernandobarraza@javerianacali.edu.co
DOI: https://doi.org/10.5281/zenodo.17204707

Overview

This deposit provides preprocessed, analysis-ready inputs used in the paper on Local pluralistic homophily and community boundaries. It standardizes several real-world networks with overlapping communities to a common format compatible with our open-source pipeline.

Included datasets

Core datasets (analyzed in the article)

• DBLP — co-authorship; communities = publication venues (conferences/journals).

• LiveJournal (LJ) — friendships; communities = user-created interest groups.

• YouTube — friendships; communities = subscription-based groups.

• StackOverflow (SO) — user interactions; communities = question tags.

• Deezer — friendships; communities = music-genre labels.

• GitHub — co-participation; communities = repository topics/languages.

• Amazon — co-purchase; communities = product-category taxonomy.

• Twitch — interaction/affinity; communities = channels or game categories.

Extended datasets (processed but not analyzed in the article)

• Orkut — user-defined groups.

• PPI (protein–protein interaction), DDI (drug–drug interaction), C. elegans neural connectome.

These extended resources are provided to facilitate replication and follow-up studies; they are not part of the main empirical analyses.

File formats

• Edgelist (*_network.txt): space-separated pairs u v; undirected; no self-loops; duplicates removed; 0-based contiguous node IDs as strings.

• Communities (*_communities.txt): each line lists node IDs in one community (TAB-separated); overlaps allowed.

• Node→communities map (*_node_to_communities.csv): variable-length rows node_id,comm1,comm2,…; empty memberships are rare but permitted.

Reproducibility

• Code repository: https://github.com/fernandobarraza/local-pluralistic-homophily
  (scripts for standardization, scoring, and evaluation; parameters and seeds documented).

• Please cite this dataset using the DOI above. If multiple versions exist, the concept DOI resolves to the latest version.

Notes

• LFR benchmarks were generated with the official LFR generator by Lancichinetti–Fortunato–Radicchi.

• Community sets are built as connected components within each group/label to preserve overlap (components of size ≥ 3).

Changelog (this version)

• Added Amazon to core datasets.

• Moved GitHub, Deezer, Twitch from extended to core.

• Kept Orkut as extended; added PPI, DDI, and C. elegans as extended (processed/not analyzed).

• Minor text clean-ups.