W-Apriori algorithm rules are not displayed

akankshatrivedi

hi,

I have taken deals default data set and converted the values into numerical to binomial then connected it with W-Apriori but its output is not displaying the rules its just telling that 1 instance is found at support 0 and minimum metric 0.9 but not showing the output. with which association box I should link my apriori algorithm so that it can display the output?

sgenzer

hello @akankshatrivedi - yes as @Thomas_Ott said this is not your error but the fact that the Weka extension is not native to RapidMiner, and hence is not always 100% compatible with the latest version.  I am tracking down the status of the Weka extension today for you.  If you really are under the gun, I would suggest trying to downgrade your version of RapidMiner (you can download any version of RapidMiner from the download page - scroll to the very bottom).  If you're not doing anything fancy, the older versions should work fine.

 

I appreciate your patience.

 

Scott

sgenzer

hello @akankshatrivedi - welcome to the community.  The W-Apriori operator is part of the Weka Extension and hence its results are rather hit-or-miss.  I will look into it a bit but FYI always encourage users to try using RapidMiner-native operators first whenever possible.

 

Scott

akankshatrivedi

hi sgenzer,

I have tried many rapid operations but having problem with this one. Waiting for your reply.

Thomas_Ott

Methinks, in general, that the Weka extension needs to be updated. 

akankshatrivedi

No @Thomas_Ott I have just updated WEKA extension, i just need to know what operator I should connect it with so that association rule will be displayed as in FP algorithm "create association rules" operator is linked to generate frequent rules.

Thomas_Ott

I'm referring to the fact that the Weka extension needs to be 'refreshed' and updated as an extension by RapidMiner, it's a few versions behind what is available. 

 

That said, there are some interoperability issues between using RapidMiner generated Weka models and trying to 'score' them in RapidMiner. This might be a similar case when trying to extract the W-Apirori rules. 

akankshatrivedi

oh!

WEKA is 7.3 and Rapidminer is 7.6 now what should I do?

I'm working on this for my project. I have to implement it by any ways.

akankshatrivedi

hello @sgenzer,

thanks for the advice but i'm searching for the older version which can support wwindows 32bit but I could not find it.

sgenzer

did you go to the downloads page and click on the "older versions" menu?  I just went there and was able to find it:

 

akankshatrivedi

please give the link, because what I can get on the site is version 7.6 not 7.3

David_A

Hi @akankshatrivedi ,

 

as far as I can see, the W-Apriori Operator is still working fine. Just by chance your combination of the ExampleSet and the default parameters of the Operator resulted in a set of empty rules.

If you change the max value in Numerical to Binominal to for example 50, the values for 'Age' are not all true. Likewise change the value of T for the W-Apriori Operator to 1.0, then you actually get two rules in your result overview.

 

The actual draw-back of the Weka Operators is, that the output is not compatible with the Apply Association Rules Operator. So I recommend to use the core Operators found under 'Modeling -> Associations' in the Operators tree.

 

I hoped that helped you.

 

Best,

David

akankshatrivedi

no I can't see the older version tab

akankshatrivedi

hey @David_A,

I have already changed the MAX values but W-Apriori is not running on iris dataset. even i have tried T=1.0 aslo but doesn't wrok. but I have taken supermarket data set and changed the values "nominal to numerical" and then "numerical to binomial", it produced the rules though it is showing the error "potential problem detected". without nominal to numerical conversion it is showing "no rules found".