How to find the best variables to use for segmentation in SAS VA?

[Babloo]

In SAS Visual Analytics I want to run variable explorations against the entire customer database, plus external data, to find the best variables to use in models for the next best offer in support of your segmentation, retention and customer value objectives.

Appreciate if someone of you provide me insights for the same.

[PetriRoine]

Hello @Babloo 

I suppose you are doing data exploration in SAS VA to find the best candidates and the actual model somewhere else?

I would utilize f.ex.:

- decision tree to understand significant variables

- histogram to check do distributions to assess need for variable transformations

- bar chart to check do I have imbalanced classes

- scatter plot and correlation to evaluate relationships in data

- correlation to check if independent variables correlate too much

- box plot f.ex. outlier detection

If you have SAS Visual Analytics version 8.3 available you can utilize Automated Analysis object to get a good view on the most significant variables per set response variable.

If you have SAS Visual Statistics or Visual Data Mining and Machine Learning available you can use the modelling objects available and see directly in the object the variable importance diagram.

I hope this helps a little.

Best regards,

Petri Roine