Accuracy sounds like the only score that matters, until your model never finds a single case Continue reading on Medium »