topic Re: Is it possible to use SAS EM to find out rules that are toggled by specified discriminant factor in SAS Data Science

Is it possible to use SAS EM to find out rules that are toggled by specified discriminant factors?

youyunjupiter — Mon, 21 Dec 2015 09:37:04 GMT

Seqecmpgroup function in R TraMineR packages is quite helpful to check the rules that are particularly related to some discriminant factors, for example, some rules might be quite different from transactions related to gender. I am wondering whether there is a quicker way to check it using SAS EM? Plan B is to run levels in those factors 1 by 1 and compare the results..

Thanks!

Re: Is it possible to use SAS EM to find out rules that are toggled by specified discriminant factor

M_Maldonado — Mon, 21 Dec 2015 14:24:52 GMT

Hi Youyun,

First time I hear about this Trajectory Miner package. From a quick look to their doc, that seqecmpgroup function identifies discrimiating subsequences using chisquare. Once with the sequences, you can use a decision tree with Chi-square criterion to confirm if certain ruels are specific to certain levels of a factor.

One way to do it would be to use the Association or the Market Basket Analysis nodes, then use a decision tree on the rules.

I hope this helps!

-M

Re: Is it possible to use SAS EM to find out rules that are toggled by specified discriminant factor

youyunjupiter — Tue, 22 Dec 2015 08:12:41 GMT

Hi Miguel, do you mean to do decision tree on scores from AA/MBA, i.e. score ~ function(factors)? I guess the key difference between this way and R function is that R function is able to calculate confidence/lift seperately for different level of factors. However, in SAS EM, it seems difficult to do, especially when start group node doesnt support transaction data. Not sure if I understand correctly...