DBSCAN (Density-based spatial clustering of applications with noise) is used to cluster data where the shape of the clusters is not important; unlike in k-means, DBSCAN clusters do not need to be spherical.
DBSCAN has two hyperparameters, ε and minPt. DBSCAN starts by looking for data points that have at least minPt other data points within a radius ε. Such data points naturally bunch together to form the clusters DBSCAN discovers. It then goes on to add any remaining data points that are within distance ε of a cluster to that cluster; the attribution is random where a data point is a potential member of two or more clusters. Any data points that then remain are marked as unclassified.
Note that the number of clusters is not one of the hyperparameters that needs to be specified.
- Density-based spatial clustering of applications with noise
- has functional building block
- has input data type
- IDT_Vector of quantitative variables
- has internal model
- has output data type
- has learning style
- has parametricity
- PRM_Nonparametric with hyperparameter(s)
- has relevance
- sometimes supports
- mathematically similar to