The error calculated by finding the difference between an actual data point and the value that a model for the data predicts.