Convert Class Probabilities of a multiclass model to scores in range 0-100

问题

What I want to do is to generate a score of 0-100 based on the predictions of a three class classification model. For eg. The predict_proba of a 3 class logistic regression model gives me 3 probabilities x, y, z as shown below -

0 1 2

x y z

Now, I want to generate a score of 0-100 based on these probabilities, where 0 is closer to class 0 and 100 is closer to class 2.

回答1:

Try this:

prob['P']=(prob['1']*1+prob['2']*2)/2

prob['0'] is multiplied by 0, so you don't need it.

examples:

prob['0']=0.5, prob['1']=0.5, prob['2']=0==>prob['P']=0.25

prob['0']=0.75, prob['1']=0.25, prob['2']=0==>prob['P']=0.125

prob['0']=0.1, prob['1']=0.2, prob['2']=0.7==>prob['P']=0.8

prob['0']=0, prob['1']=0, prob['2']=1==>prob['P']=1

来源：https://stackoverflow.com/questions/57577489/convert-class-probabilities-of-a-multiclass-model-to-scores-in-range-0-100