For the model, the most important thing is two things
Precision: the model output is the number of ai that is really ai / the number of ai that is output in the model
Recall rate: the number of output ai in the model / the number of real ai
Then I want to judge the extreme value and variance of athletes' performance when I was a child
We sometimes fail to judge how good a model is
Then we have to use a single indicator to evaluate the quality of the model
Let the precision be a and the recall be b
We use the harmonic mean of a, b to evaluate the model
As for why, I don't know
F1 = 1 / ((1 / a) + (1 / b) )
= (a * b) / (a + b)
There may be other factors that affect the quality of your model
You can design an indicator yourself, such as weighted average