Feature Selection

andre_marquesandre_marques Member Posts: 2 Contributor I
edited November 2018 in Help
Hello.
I'm doing a work where i need experiment several methods of Feature Selection in rapid miner and then compare them. I'm testing rapid miner to do this work but i don't be well succeed with some operators. One of them are ANOVA. What i have to do to use use this operator.  ???
For example in the operator CFSFeatureSetEvaluator i use this inside another operator (FeatureSelection) but with ANOVA i don't now what to do ???
Someone can help?

Regards,

André.

Answers

  • IngoRMIngoRM Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder
    Hi,

    For example in the operator CFSFeatureSetEvaluator i use this inside another operator (FeatureSelection) but with ANOVA i don't now what to do
    The first part is perfectly fine - but I don't get why you would want to integrate the ANOVA operator inside this setting?

    The ANOVA operator can only used to calculate if two performance vectors (which in fact could guide the feature selection process) significantly differ.

    Cheers,
    Ingo
  • andre_marquesandre_marques Member Posts: 2 Contributor I
    My idea is not integrate the ANOVA operator inside these setting but test separately ANOVA operator and CFSFeatureSetEvaluator in a data and then observe and conclude how can be better using for example SimpleValidation. I don't now if i am explicit... Basically my question is what operator i can apply to my data before use ANOVA  to do work this operator...

    Cheers,

    André.
  • IngoRMIngoRM Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder
    You cannot use ANOVA as a performance calculator which could then be used to guide the feature selection - this is simply not what ANOVA does. As I said before, ANOVA can only be used for comparing two performances with respect to the question if they significanctly differ. But ANOVA cannot be used to create a performance measure.

    You could get an idea of what can be done with ANOVA by looking into the samples delivered together with RapidMiner. There is also a sample process showing how ANOVA can be used.

    Cheers,
    Ingo
Sign In or Register to comment.