DBSCAN (Density Based Clustering of Applications with Noise)

Publication Year: 1996

Access: Open

Link: https://cdn.aaai.org/KDD/1996/KDD96-037.pdf?source=post_page

Author(s): Ester, M., Kriegel, H.P., Sander, J., and Xu, X.

Organisation(s)/Authors: University of Munich

Description:

Clustering algorithms are used for class identification in spatial databases. The Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is designed to identify clusters with arbitrary shapes, specifically for large datasets. It relies on a "density-based notion of clusters which is designed to discover clusters of arbitrary shape" (Ester et al., 1996). DBSCAN requires only one input parameter and supports the user in determining an appropriate value for it. Tilloy et al., 2022 has conducted a more recent application of DBSCAN for the evaluation of multi-hazards, which can be found at the following link: https://esd.copernicus.org/articles/13/993/2022/esd-13-993-2022.html

Technical Considerations:

R package(s): https://cran.r-project.org/web/packages/dbscan/index.html python package: https://pypi.org/project/dbscan/

Key Words:

spatiotemporal clustering, multi-hazard modeling, hazard footprints