10-16-2011 12:11 PM
This might have a simple answer but i just want to be sure.
Lets say that through clustering i have created 6 clusters , each cluster containing various IDs whith its characteristics lets say like up to 5 attributes (columns for each ID).
Since i have proven my model to be robust based on historical data my question is the following:
When i receive lets say 10000 new IDs (with their 5 attributes) and i want to place them into the clusters i have already created given based on their similar attributes , what would be the best procedure to follow?
I hope this makes sense , thnx in advance
10-16-2011 03:14 PM
One way is to compute the center of each cluster (often the mean of the cluster is used). For each new observation, compute the distance to each cluster center, and assign it to the cluster to which it is closest. (If the overall data has covariance, then you could use a Mahalanobis distance instead of a Euclidean distance.) To implement this approach, look at http://support.sas.com/kb/30/662.html, which is similar in spirit.