How can I classify one example into multiple classes if necessary?

Prentice

Hello,

I have yet another problem/something I don't know. I've looked on similar cases but couldn't find an answer for my case.

How can I make test data with multiple labels?
How can I classify one example into multiple classes if necessary?

How can I make test data with multiple labels?
Say I have an example:
我喜欢香蕉和橘子。
The labels for this sentence are "banana" and "orange".
Would the input be another attribute with label 2, so that I have an attribute label 1, label 2 and maybe even label 3? Or the example twice but then with a different label, all under the same label attribute?
我喜欢香蕉和橘子。- >香蕉
我喜欢香蕉和橘子。-> orange

How can I classify one example into multiple classes if necessary?
I know that Polynomial to Binomial Classification is a thing, but I don't understand what it does after reading the help or the tutorial process.
The problem is also that not every example necessarily has two labels, a lot have one, some have two and a few could have three.
I thought of something like this, but I don't know if it would work or if it's effective.
For a case with an example that has two classes:
我喜欢香蕉和橘子。The probability would probably be something like Banana:0.435, Orange: 0.422, Pear:0.093, Apple:0.05.
And then for Probability<0.5 (or between 0.4 and 0.7, something like that) the two highest are the two designated classes. If higher probability higher than 0.5 (or 0.6 or something) it's one class.
The same would be for a case with three classes, but then the probability needs to be around 0.33.

Or am I thinking too difficult now? I don't know, and that's why I need your help.

Thanks in advance
普伦蒂斯

IngoRM

Hi,

There is an operator which can create a desired number of predictions based on the confidence values of a model. It is called "Generate Prediction Ranking". Below is the XML of a little example process.

Hope this helps,
Ingo

<？xml version = " 1.0 " encoding = " utf - 8 " ?> <过程版本sion="9.2.001">
  
    
    

    
  
  
    
    
    
    
    
    
    
      
        
      
      
        
          
          
        
        
        
        
      
      
        
      
      
        
        
      
      
        
        
      
      
      
      
      
      
      
      
      
      
    
  

MartinLiebig

Maybe you want to use Generate Prediction Ranking?

Best,

Martin

Prentice

Thanks, I'll try to make something out of this.