36229_en-US

liliconcom

Cluster a mixture of continuous and catergorical variables

I have a customer that wants to cluster a mixture of numeric and catergorical variables.  I suggested binning the continous data and using latent class analysis.  Does anyone else have other suggestions or techniques for this type of clustering.

 

Thanks,

 

Peter Hersh

Re: Cluster a mixture of continuous and catergorical variables

Here's a journal with a quick example - lots of great visualizations. You have to practice with the magnifier to read the lables, just fyi. Pivot on selected clusters is helpful too for moving things around. 

Re: Cluster a mixture of continuous and catergorical variables

Hierarchical Clustering takes both categorial and continous data for clustering. Turn on 2-way clustering for the best view for interpreation across variables. 

Re: Cluster a mixture of continuous and catergorical variables

Hey Peter,

 

Good suggestion.

 

It sounded like multiple catergorical factors mixed with a couple continous. 

Re: Cluster a mixture of continuous and catergorical variables

Pete:

 

How many levels for the categorical factor? If it's just a few and the levels might make practical sense...to remain differentiated, perhaps using a 'By' statement for each level could work too? I'm thinking something like residents of states? Where a mixing together people in various states might not make sense from a practical point of view depending on the problem at hand?