Cluster algorithm for coordinate based clustering with revenue density

55 Views Asked by At

I am looking to divide an area into n clusters. The dataset consists of stores with coordinates and their revenue. However I do not know which clustering algorithm I should use.

After some research I always end up with K means, which does an okay job with just the coordinates as features, even with Euclidean distances, however, of course when I add another non location based feature the centroids are not location centroids.

To include revenue as a feature I have multiplied the store data points by a revenue based factor, to increase point density in an area with a higher relative area.

Should I try to redistribute the Kmeans data points to create evenly sized cluster?(Even similar sized clusters based on amount of stores only would also already help a lot) If so how, because then Euclidean distance doesn't make sense anymore to assign it to the nearest cluster.

Or should I use something like DBSCAN, which I've tried, but not managed to fit to my coordinate data yet.

0

There are 0 best solutions below