voting with weight factor

s_ghorbanis_ghorbani Member Posts: 2 Learner I
edited July 2019 in Help
Warm hello
I am going to use some algorithms for prediction by voting because of promotion of accuracy absolutely I want to give a weight factor to every operator based on it's accuracy in independent prediction. Please guide me how can I do it.

Best regards
Saeed
Tagged:

Answers

  • MartinLiebigMartinLiebig Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 3,503 RM Data Scientist
    do you want to set these weighting factors manually or let them be determined by an algorithm?
    Cheers,
    Martin
    - Sr. Director Data Solutions, Altair RapidMiner -
    Dortmund, Germany
  • s_ghorbanis_ghorbani Member Posts: 2 Learner I
    Dear Martin
    So thanks.
    I want to set these factors automatically by algorithms. 
    Cheers
    Seed
  • PapadPapad Member Posts: 68 Guru
    @mschmitz
    Can you please answer for both situations?
    Thanks in advance.
  • Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,635 Unicorn
    I'm interested in this as well.  There is an Vote (Ensemble) algorithm that allows you to easily combine predictions from multiple ML operators, but this simply uses majority voting---it does not allow you to weight the combined prediction in any way.  I've already suggested the addition of "weight by confidence" to the Vote operator, which would be helpful at the individual example level, but what you are asking about it is weighting at the overall operator level based on some global measure of performance, presumably.  I don't know of any easy way to accomplish this in RapidMiner (I can imagine how it could be done but very manually using Generate Attributes to create your own weights and then average your predictions by using those weights).
    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
Sign In or Register to comment.