Link sign prediction

2 datasets tagged as the tag “Link sign prediction”.

2 datasets.

Epinions Social Network (Signed) Analyzed

A Epinions.com (consumer review platform) dataset of 131,828 users and 841,372 directed signed edges (717,667 trust, 123,705 distrust), collected by Richardson et al. via Epinions.com crawl, distributed by SNAP. Each record contains directed signed edges between users: +1 indicates explicit trust, -1 indicates explicit distrust, capturing both positive and negative social relationships. one of the largest publicly available signed social networks, with explicit user-declared trust and distrust labels rather than inferred sentiment — making it ideal for signed graph algorithm benchmarking. Suitable for signed network analysis, trust propagation, link sign prediction, community detection in signed graphs, and balance theory validation.

131.8K records131.8K nodes841.4K edges2003

Slashdot Zoo Signed Network (Cumulative) Analyzed

A Slashdot.org (technology news and discussion) dataset of 82,144 users and 1,601,229 directed signed edges (1,234,967 friend, 366,262 foe) merged from 3 snapshots, collected by Leskovec et al. via Slashdot crawls over November 2008 to February 2009, distributed by SNAP. Each record contains directed signed edges between users: +1 indicates an explicit friend tag, -1 indicates an explicit foe tag; three temporal snapshots merged into one cumulative graph. one of the canonical signed social network benchmarks, notable for its explicit friend/foe labeling system and use as a primary dataset in foundational signed network research including balance theory and status theory studies. Suitable for signed network analysis, link sign prediction, structural balance theory, status theory validation, community detection in signed graphs, and trust/distrust propagation modeling.

82.1K records82.1K nodes1.6M edges2008