2

Can anybody tell me the formula how to find the number of false positives with respect to the first class?

enter image description here

where $y$ is the truth / target and $a(x)$ is the prediction

Louis T
  • 1,148
  • 8
  • 22
Bootuz
  • 121
  • 2
  • It would be useful if you posted what $y=i$ and $a(x)=i$ mean. I am not sure what exactly is happening here. Is this a classification problem? – learning Nov 07 '17 at 22:54
  • @learning it is a confusion matrix with three classes – Bootuz Nov 07 '17 at 23:22
  • So if the column labels indicate the predicted labels and the row labels indicate true labels, then false positives for the first class would be the ones whose true class is 2 or 3, but classified into the first column. – learning Nov 07 '17 at 23:30
  • So row 2 and row 3, column 1 indicate the ones that have different true labels than the ones predicted right? So it should be 5+10. – learning Nov 07 '17 at 23:35
  • it says that's a wrong answer.. – Bootuz Nov 07 '17 at 23:36
  • Then maybe I don't know what the true labels are and what the predicted labels are. – learning Nov 07 '17 at 23:37

1 Answers1

0

I always find the notion of false positive and negative confusing, especially when it comes to multi-class problems.

A good rule of thumb that I came up with is the following

True positive: "I predicted that you were a certain class, and I was right"

False positive: "I predicted that you were a certain class but I was wrong"

True negative: "I predicted that you weren't a certain class, and I was right"

False negative: "I predicted that you weren't a certain class, but I was wrong"

So in your case, there are 31 true positives with regard to the first class since you predicted 31 times that something was class 1 but you were wrong.

Valentin Calomme
  • 6,026
  • 3
  • 21
  • 52