1

I encountered following question and its answer but I do not know the justification.

Original Question:

Given an imbalanced data, 90% -ve and 10% +ve. When the cost of wrongly predicting Positive point as negative is very high, mainly in medical domain.

Which performance metric shall be used?

Answer given is, F1-score.

My question is,

Why F1 score is appropriate in this case?

Ajey
  • 155
  • 4
  • Someone may be able to explain the thinking behind recommending the F1 score in this case. I'll say that it's still wrong. As is all hard 0-1 classification, *especially* in a medical context. Background at https://stats.stackexchange.com/a/312124/1352 and https://stats.stackexchange.com/a/312787/1352 (every criticism of accuracy as a KPI at the second link applies equally to F1). – Stephan Kolassa Jul 08 '20 at 12:31

0 Answers0