Why confidence value is infinity?

kim_joa

I found some problem in FP-Growth operator. 
The FP-growth has a parameter named "min items per itemset". It is set default value as 1.
In my process  I set  the value as greater than 1, in result I received some rules that confidence value is infinity.
I did so because I didn't need the itemsets of size 1. 
I am sure there is no relation between confidence and that parameter value.
I think it might be bug.
I hope rapidminer developer to verify and to improve the problem.
thanks.

yyhuang

Confidence is support divided by another support. from wiki 
 
if we mistakenly have supp(x)==0 the confidence for the association rule X=>Y will be inf.
It seams a bug to me.
@mschmitz may have some ideas https://github.com/rapidminer/rapidminer-5/blob/master/src/com/rapidminer/operator/learner/associations/AssociationRuleGenerator.java

sgenzer

@yyhuang @kim_joa I'm happy to push this to Product Feedback but please note that the GitHub link you listed is from RapidMiner 5.x. Can you please confirm that the bug still exists with our most recent version? https://github.com/rapidminer/rapidminer-studio/blob/master/src/main/java/com/rapidminer/operator/learner/associations/AssociationRuleGenerator.java

Scott

lionelderkrikor

Hi @sgenzer,

I confirm that this bug is still present in RM 9.2 : 



Regards,

Lionel

IgorE

The bug is still present in RM 9.6

sgenzer

hi @IgorE yes we know the issue is still present in 9.6. That's why I wrote in the status that as of 9 Mar 2020 Engineering was still working on it. It will likely be addressed in 9.7 but I cannot promise anything. Please watch the status headers as that is always your best source of information.

Scott