Machine Learning in Java - Second Edition

Mean squared error

Mean squared error (MSE) is an average of the squared difference between the predicted and true values, as follows:

The measure is very sensitive to the outliers, for example, 99 exact predictions and 1 prediction off by 10 is scored the same as all predictions wrong by 1. Moreover, the measure is sensitive to the mean. Therefore, a relative squared error that compares the MSE of our predictor to the MSE of the mean predictor (which always predicts the mean value) is often used instead.