DESCRIPTION:
A set of algorithms based on te collective factorization on related matrices approach. Spectral Relational Clustering discovers hidden structures of mutli-types of objects. Makes simultaneous use of both feature and relation information. Illustrative applications include analytics, market basket shopping patterns, and detection of terrorist associations.
ADVANTAGES:
- Works with real-world data sets involving objects of multiple types
- Reveals not only local but global hidden structures
PATENT STATUS:
Patent Pending |
