View Full Version : MPCluster: Find Clusters in Pushpin Data
02-11-2010, 08:53 AM
We have a new product that is currently in beta testing. Called "MPCluster" it can find clusters of data points in MapPoint pushpin data.
For further information, see the provisional website,here:
MPCluster: Cluster Analysis for Microsoft MapPoint (http://www.mpcluster.com)
07-12-2010, 04:45 PM
I downloaded your software a few weeks ago and started playing around with it. Nice work! The program does a fine job of clustering large amounts (over 1000) of pushpins together. I tried clustering a massive dataset into groups of 10,20,50 and 100 with good results.
One area of the program that I would like to suggest for improvement is the "Unallocated" output of the program. Usually the pushpins that fell into this output were the statistical outliers that really didn't cluster well with anything. If there is a way to eliminate the "Unallocated" output of the program and instead definitively assign those pushpins to clusters, that would be extraordinarily helpful.
Also, I know you probably won't share this since it pertains to the inner workings of your program, but I'll ask anyway :). I noticed that each time I run the program it takes roughly the same amount of time to compute regardless of the amount of pushpins I am clustering. Did you use a grid layout as part of your initial heuristic approach?
07-13-2010, 08:48 AM
Thanks for the suggestion about outliers - I've made note, although no promises. At the moment, you could change the cluster size limits, although these would also change the clusters themselves (to get that "tiling" effect you can see in some of the examples).
A grid layout would not be appropriate as clusters are rarely in a grid.
Powered by vBulletin® Version 4.2.0 Copyright © 2013 vBulletin Solutions, Inc. All rights reserved.
Search Engine Friendly URLs by vBSEO