"FPGrowth/Assiciationrule Generator - filter useless rules"

wernerwerner Member Posts: 9 Contributor II
edited June 2019 in Help
Hi,

I am currently testing the FPGrowth algorithm using an example set which is structured like this

id          attribute1    attribute2    attribute3
-----------------------------------------------------
4711    true              false            true
4712    false            false            false
...

Now, FPGrowth followed by AssociationRuleGenerator generates rules like

IF attribute1 = 0 THEN attribute2 => 0

These rules are not important (at least for me), because they just say that if attribute1 does not appear, attribute2 does not appear.
Is there any other way to filter such rules from the FrequentItemset? I tested the FPGrowth-Parameter "must contain" with the simple regex "attribute1 = 1", but it had no effect

Thanks for your help !
Werner

Answers

  • wernerwerner Member Posts: 9 Contributor II
    I think I solved it. I used the parameter "positive value" in FPGrowth and set it (for whatever reason!!) to "false".
    Now it produces only rules which start with
    IF attribute = 1 THEN ....
  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,531 Unicorn
    Hi,
    but the generated rules do not rely on the correct data. Instead you are using the inverted data now: Each false is treated as being true. In normal transaction data, the false case is much more frequent than the true case. So rules based on the false case will have a greater support. That's why you find some rules and don't when switching back to "true". You might lower the support for finding rules, if there aren't any found.

    Greetings,
      Sebastian
Sign In or Register to comment.