-2

can you please tell me what does this result of prediction mean ? actually it is supposed to show yes or no !

{"probabilities": [0.9990405440330505, 0.0009594644070602953], "class_ids": [0], "classes": ["0"], "logits": [-6.948175430297852], "logistic": [0.0009594644070602953]}

Sofia
  • 29
  • 5

1 Answers1

1

Guessing:

It looks like you have a binary classification problem (two classes, class_id=0 and class_id=1). The probabilities definitely indicate Prob(class_id=0) = 0.999 thus the prediction from classes is also 0.

Then depending on how you define "yes"/"no" (which one is class 0 and which one is class 1), you should have the answer.

greeness
  • 15,956
  • 5
  • 50
  • 80
  • Thank you for your answer @greeness , actually I don't understand why the result is class 0 or 1 , I didn't define this in my model, my labels are yes / no , and before that it was >50k , =<50k , and the result was >50k and =<50k. Than I don't know why when I changed the labels , I have these results instead of yes , no – Sofia May 18 '18 at 07:43
  • "yes/ no" is a mutual exclusive two-class problem. Usually "no" label is used as class 0. If it is the case, then you can just interpret the result like this: "if classes is 0, report "NO"; otherwise, report 'YES'". I am not familiar with cloud ml api; maybe they have update on how to report the result. – greeness May 18 '18 at 17:07