02-22-2016 04:17 AM
I am making Customer segmentation. My data included numeric data (4 variables) and categories data (7 variables).
I used clustering method to classified customer.
If I transfer categories data to dummy and then use K-Means algorithms for new data, It make falsifying clustering result or not?
Could you advice to me the way that I can use to cluster for both Numeric and Categories data on SAS?
Thanks for your help!
02-23-2016 09:17 PM
Clustering methods in EM convert nominal variables into dummy variables. There are different encoding techniques available in the cluster node. You can try those options and check the compactness of clusters. If you are using procs, you can also try PROC CLUSTER by using distance/similarity measures (such as jaccard).