Accuracy attempts to measure how well a classifier performs in general, giving equal weight to positive and negative predictive value.
However, in the presence of class imbalance, accuracy can give a false sense of confidence. For example, if positive cases are extremely rare, then a classifier that always predicts in the negative will appear to be highly “accurate.”
F-scores can address the problem of class imbalance. However, they are focused on the positive class, and hence do not provide a direct means of rewarding negative predictive value.