# Actor Co-occurrence Network

**Task**: Actor category classification  
**Size band**: medium · **Label type**: semantic  
**Label column**: `actor_class` · **Converter**: `converters/convert_geomgcn.py`

Actor-only induced subgraph of a film-director-actor-writer network; nodes are actors, edges are Wikipedia page co-occurrence, labels are 5 actor categories. Heterophilous benchmark (a.k.a. 'film').

## Converted graphs (neext/)

| graph | nodes | edges | classes | feature cols | isolated | class counts |
|---|---|---|---|---|---|---|
| default | 7,600 | 26,659 | 5 | 932 | 0 | 4: 1,965, 3: 1,815, 2: 1,630, 1: 1,337, 0: 853 |

*Conversion notes*: geom-gcn film out1 files; sparse keyword indices one-hot expanded to f_* columns; 5 actor classes; co-occurrence edges symmetrized.

## Source

- [out1_graph_edges.txt](https://raw.githubusercontent.com/graphdml-uiuc-jlu/geom-gcn/master/new_data/film/out1_graph_edges.txt) — 323,820 bytes, sha256 `997cdce67ea9c292…`, fetched 2026-07-23
- [out1_node_feature_label.txt](https://raw.githubusercontent.com/graphdml-uiuc-jlu/geom-gcn/master/new_data/film/out1_node_feature_label.txt) — 207,718 bytes, sha256 `1d5cdd094f8b91e7…`, fetched 2026-07-23

**License**: Not stated (geom-gcn repo)  
**Citation**: Tang, Sun, Wang, Yang. Social Influence Analysis in Large-scale Networks. KDD 2009 (origin); Pei et al. ICLR 2020 (packaging).

- https://github.com/graphdml-uiuc-jlu/geom-gcn

## Caveats

- Sparse keyword indices one-hot expanded to 932 f_* columns.

---
*Generated by converters/make_cards.py; stats from metadata.json. Raw files: `source/`. NEExT tables: `neext/`.*