# creating clusters of point data

I have a map of stores for a retailer. I am trying to form clusters (of these stores) such that each member of a cluster is at most 5km away from any other member and any non-member is at least 10 km from any member. I also would like to create a field in the point dataset indicating to which cluster a specific store beongs to. What would be the easiest way to do this? Thanks

• This problem needs to be framed differently, because its objectives can be contradictory. E.g., what would be a solution for two stores that are 8 km apart? They can't be part of the same cluster but they can't be put into separate clusters. Oct 31, 2011 at 19:37
• maybe single stores 8km apart are in their own group. Oct 31, 2011 at 19:54
• @Brad Putting stores 8 km apart into a group violates the first criterion. I gave only the simplest example of the internal contradictions. What would be done with 3 stores spaced 4 km apart on a line, for instance? Or with a bunch of stores spaced along the perimeter of a 4 km radius circle around another store? Oct 31, 2011 at 21:28
• umut, IMHO it's a good idea to sketch (with paper and pencil) the clustering criteria, that will certainly help you to reach a better approach. Also, look for some clustering theory (en.wikipedia.org/wiki/Cluster_analysis) to understand the machine's interpretation of clusters. Nov 1, 2011 at 15:16