---
configs:
- config_name: hero_hero_comic
  data_files: hero_hero_comic.csv
- config_name: adjacency_list
  data_files: adjacency_list.csv
- config_name: adjacency_matrix(preview)
  data_files: processing_data/adjacency_matrix_preview.csv
license: afl-3.0
language:
- en
size_categories:
- 10K<n<100K
task_categories:
- feature-extraction
---

# Dataset Card for Marvel Network

This is a dataset for Marvel universe social network, which contains the relationships between Marvel heroes.

## Dataset Description

The Marvel Comics character collaboration graph was originally constructed by Cesc Rosselló, Ricardo Alberich, and Joe Miro from the University of the Balearic Islands. They compare the characteristics of this universe to real-world collaboration networks, such as the Hollywood network, or the one created by scientists who work together in producing research papers. The dataset contains heroes and comics, and the relationship between them got from Marvel comics

This dataset consists of three subsets, hero_hero_comic.csv, adjacency_list.csv and adjacency_matrix.csv. 

- **Language(s) (NLP):** [EN]
- **License:** afl-3.0

## Uses

**Network Topology Analysis,** 
Just like what the original work did, we can try to analyze the network topology to discover whether the interpersonal relationships and social structures created in novels conform to real-world social networks, whether they fit the small-world model; we can also use this to infer the core heroes of the Marvel Universe, etc.

**Community Detection,**
This is an excellent toy model for applying community detection algorithms, and we naturally have clear community outcomes such as the "Avengers Alliance" for comparison.

**Link Prediction, Attack Robustness, etc.**
In fact, all algorithms effective for conventional network models can be transferred and applied to this network, including but not limited to: link prediction: which predicts which heroes might collaborate in future works; attack robustness: studying whether Thanos's snap would damage the hero ecosystem of the entire Marvel Universe; word2vec: assigning network structural features, finding heroes paires like Spider-Man and Venom...


## Dataset Structure

**hero_hero_comic.csv:**
A table three columns hero1[str], hero2[str] and comic[str]. hero1 and hero2 shows together in a same comic, which is the comic in the third colunm. 

**adjacency_list.csv:**
A table contains three columns hero1[str], hero2[str] and counts[int], which is the total times hero1 and hero2 shows together in a comic.

**adjacency_matrix.csv:**
A matrix contains 6267 columns times 6267 rows, each cell with an [int] value. Each row/column represents a hero, each nondiagonal cell represents the total times the colunm hero and the row hero shows in a same comic and each diagonal cell represents the total counts of the colunm(row) hero in this dataset.
**This file cannot be accessed by config directly, please download by filename.**

## Source Data

The sourse data is from https://www.kaggle.com/datasets/csanhueza/the-marvel-universe-social-network, which is collected from Marvel comics.

## Source Data Producers

The sourse data is created by Claudio Sanhueza

## Citation

Alberich, R., Miro-Julia, J., & Rosselló, F. (2002). Marvel Universe looks almost like a real social network. arXiv preprint cond-mat/0202174.

## Dataset Card Contact

Haocheng Qin, Duke Univ.,
hq37@duke.edu