Clustering with mixed variables

biaf08 — Fri, 05 Oct 2018 08:17:06 GMT

Hi,

I am fairly new to SAS and could use some help. I have been working on customer segmentation on a 30 000 entries data set with mixed variables (continuos, binary and categorical).

I initially was thinking of turning some of the categorical variables with few levels into binary variables and then use PROC FASTCLUS on it. However it seems like FASTCLUS only performs k-means, which is not appropriate for binary variables. I then used PROC DISTANCE to create a gower's distance matrix directly for the mixed variables data set to feed into PROC CLUSTER. But now I am getting an error and warning:

WARNING: Unable to allocate sufficient memory. Amount requested was 0, amount available was 1691620352...

ERROR: Invalid position -2147479016 for utility file WORK.'SASTMP-000000029'n.UTILITY

Am I doing something wrong or is my data set to large to be processed with PROC CLUSTER. Are there any alternative ways to cluster mixed variables ds?

Thanks 🙂

Re: Clustering with mixed variables

Ksharp — Fri, 05 Oct 2018 10:07:14 GMT

You could directly use PROC CLUSTER, if you already have design matrix like:

sex

-->

F M

1 0

0 1

topic Re: Clustering with mixed variables in Statistical Procedures

Clustering with mixed variables

Re: Clustering with mixed variables