Finding hidden patterns represents a key task in terrorism research. In light of this, the present work seeks to test an innovative clustering algorithm designed for multi-partite networks to find communities of terrorist groups active worldwide from 1997 to 2016. This algorithm uses Gower’s coefficient of similarity as the similarity measure to cluster perpetrators. Data include information on weapons, tactics, targets, and active regions. We show how different dimensional weighting schemes lead to different types of grouping, and we therefore concentrate on the outcomes of the unweighted algorithm to highlight interesting patterns naturally emerging from the data. We highlight that groups belonging to different ideologies actually share very common behaviors. Finally, future work directions are discussed.
|Nome||STUDIES IN COMPUTATIONAL INTELLIGENCE|
|Convegno||7th International Conference on Complex Networks and their Applications, COMPLEX NETWORKS 2018|
|Periodo||11/12/18 → 13/12/18|
- Artificial Intelligence
- Community detection
- Multi-partite networks
- Unsupervised learning