Ask HN: Working with GPS Cord to find distance and clusters.

I have been tasked with a rather interesting project, in an area where I don't really have much experience.

I was given the address of 10,000 customers and reverse geocoded (GPS Cordinates) them but now i am interested in finding clusters of customers. I really have no clue how to start other than to visually plot them on google maps api and search visually.

They are all in 1 city so they will be pretty condensed anyway but there will be clusters. How would i go about program-matically finding these clusters. I have done some google searches and nothing has really turned up.

Thanks for any help.

2 comments

[ 3.3 ms ] story [ 19.9 ms ] thread

bartonfink 14y ago ↗

What you want to look at is k-means clustering. It's a machine learning technique that will abstract to any number of vectors, but it will handle what you're looking for.

http://en.wikipedia.org/wiki/K-means_clustering

redslazer 14y ago ↗

Wow there is a lot of information about that algorithm on the web. Im sure ill be able to cobble something together from that. Cheers