2,708 scientific publications categorized into seven distinct classes, connected by a citation network of 5,429 links. Each entry includes a binary word vector representing the presence or absence of 1,433 unique dictionary terms.
Use Cases
- Train graph convolutional networks (GCNs) to predict publication categories using the citation links and word vectors
- Evaluate node classification algorithms based on the seven provided class labels
- Perform link prediction tasks to identify potential missing citations within the 5,429-link network
Strengths
- 2,708 nodes representing individual scientific publications
- 5,429 edges representing citation links between papers
- 1,433-dimensional binary feature vectors indicating word presence
- Seven mutually exclusive classification labels for each publication