Measures how often a model predicts the wrong class