Machine Learning Algorithms - Second Edition

R2 score

Another very important metric used in regressions is called the coefficient of determination or R². It measures the amount of variance of the prediction which is explained by the dataset. In other words, given the variance of the data generating process p_data, this metric is proportional to the probability of predicting new samples that actually belong to p_data. Intuitively, if the regression hyperplane approximated the majority of samples with an error below a fixed threshold, we can assume that future values will be correctly estimated. On the other side, if, for example, the slope allows to have a small error only for a part of the dataset, the probability of future wrong prediction increases because the model is not able to capture ...