"How to associate to 'others' if below a certain confidence level ?"

kaymankayman Member Posts: 368   Unicorn
edited June 11 in Help

Hi there,

 

Not too sure how to describe this best so be gentle :-)

 

I have a given example set, and through the usual ML processes I was able to get a range of 5 common association labels. The problem is that the system will associate now any new examples to any of these 5, if the confidence is high enough there is no problem, but if there is no match or little confidence it seems to associate the example at random to one of the 5 options.

 

What I would therefore like to achieve is the following : If the confidence is above a given level (say 80%) assign the example to the corresponding label, otherwise assign it to a generic label, like 'other'.

 

This way we have an easy way to improve the model (or add new labels) based on what ends up in the 'other' category.

 

Is this feasible, and if so, what would be the best way to achieve this?

Tagged:

Best Answer

  • mschmitzmschmitz Posts: 2,111  RM Data Scientist
    Solution Accepted

    Hi kayman,

     

    the operator Drop Uncertain is doing what you want.

     

    ~Martin

    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
Sign In or Register to comment.