"Association Rules"

iodean

I am new at RM and I have been going through the book Data Mining for the Masses. My first question has to do with the operator "Create Association Rules". Under 'Show rules matching' it seems to pick up the first 4 or 5 combination of attributes that come from FP Growth. Can one control which attributes should be shown in the rules?

My second question has to do with how to replace missing values - is there away to replace the missing values with a random allocation of values in a certain percentage. For example, using the Congressional Voting Records from the UCI Machine Learning Repository, is there a way to replace missing votes by a random "yes" or "no" in a proportion of say 60:40?

Thank you

MariusHelf

Heya,

please use one post for one question the next time, that helps us to keep the forum clean and structured.

Concerning the association rules, you can't define which attributes are chosen - if you would to that, you could as well create the rules by hand and make the operator obsolete. However, you can modify the threshold such that only rules with a given confidence are kept.



The Replace Missing Values operator does not allow that kind of replacements, but you can as well use Generate Attributes with an expression like this:

if(missing(attribute), if(rand() < 0.4, "yes", "no"), attribute)

Best regards,
Marius