[R] How can I compare two apriori rules created from the same dataset

Benoit Vaillant beno|t@v@|||@nt @end|ng |rom no-|og@org
Sun Sep 15 13:46:50 CEST 2019


Hello Chris,

On Sat, Sep 14, 2019 at 05:27:31PM -0300, Chris Chinedozie wrote:
> I have a dataset and I divided it into 2 datasets RANDOMLY A and B, where A
> is 70% of the main dataset and B is 30%... Then I applied Apriori algorithm
> on the both A and B separately, generating its rules..

My best guess is that you don't want to do that.

Association rules hugely rely on support for building the itemset. And
then confidence for the orientation of the rule.

Both are very simple metrics. Yet if you divide your dataset, this
will lead to very different itemsets due to the algorithm.

> I want to compare rules from dataset A to rules of dataset B

Can you precise « compare »?

> example:
> A has the following rules
> [(sex=0,age=1),(sex=1,age=1,money=0),(sex=0,age=2,money=2)]
> 
> B has the following rules
> [(sex=0,age=1,money=1),(sex=0,age=1,money=0),(sex=1,age=1,money=0)]

Aren't these itemsets rather than rules?

> A(sex=0,age=1) => B(sex=0,age=1,money=1) returns TRUE
> A(sex=1,age=1,money=0) => B(sex=1,age=1,money=0)) returns TRUE
> A(sex=0,age=2,money=2) => B(sex=1,age=1,money=0) returns FALSE

I don't see rules, nor sense.

Can you clarify?

Thanks!

-- 
Benoît Vaillant

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 866 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20190915/b102bb5d/attachment.sig>


More information about the R-help mailing list