Hi,
I have a multi label problem. i have coded the model, used BCEloss(with logits), can get the prediction from the model. But i am confused on what metric to use. I used accuracy score, but its obvious accuracy score will give good score
label = [1,0,0,0,0,0,0,0,1,0,0,0]
pred = [0,0,0,0,0,0,0,0,0,0,0,0]
This means even a 0-model(returning 0 always) gets more accuracy.
But i searched the whole internet(may be not whole), people don’t usually use metric or use accuracy score as metric.
Shouldn’t we use a metric which considers the positive labels and predictions like precision, recall…?