Clustering with DBSCAN

This workflow performs clustering of the iris dataset using DBSCAN. Notice the Numeric Distances node to feed the DBSCAN node with the matrix of the data to data distances. Distance type can be selected within the Numeric Distances node.

This is a companion discussion topic for the original entry at

Hi, I am trying to apply DBSCAN to a spatial analisys. In short: I have a series of GPS coordinates and I want to group them by DBSCAN in order to identify some ‘hot spot’ where coordinates points are more ‘dense’. As I have coordinates in format 43.345 - 11.456 (degrees decimal) the euclidean distance does not work: I should set ‘eps’ far smaller than 0,5 (the minumun accepted). One way to solve should be to calculate Haversine distance instead of Euclidean: is it possibile? If not, have you some suggestion?
Thank you so muche and all my compliments for the excellent work you are doing

1 Like