2

I have created assocociation rules by using the the Apriori Algorithmus of the arules package. Now, i would like to validate the resultes, with Validation method (e.g. cross Validation)

But i don´t find any opportunitys or function in r, which automatically apply the created rules to a test data and check the results.

I haven´t attached any sample code or data, because it´s a simple Apriori function. I would like to know, if there is basically a opportunity to validate association rules with test data?

If you need further informations, or my question is too imprecise, please let me know

Thanks in advance!

Tom Maier
  • 33
  • 5

1 Answers1

1

With association rules you have implicit validation when choosing the required support and confident values to generate the rules. You won't find any validation easy function in the arules package.

In a way, Confidence, support and lift are the validation.

Nevertheless, there are some papers tackling this issue, but they are very field specific and to develop a validation algorithm in order to eliminate non-predictive rules, you need to understand what you want to rule out depending on your particular problem...

One of them is this one that I read a while ago:
Association Rule Discovery With the Train and Test Approach for Heart Disease Prediction
Another one is this:
VALIDATION OF ASSOCIATION RULES WITH FINITE AUTOMATA AND HIERARCHICAL CLUSTERING

Felipe
  • 8,311
  • 2
  • 15
  • 31