MAE and RMSE — Which Metric is Better?

RMSE is better for grid search hyperparameter tuning (which in sci-kit-learn is based on test set score), while MAE is better for interpreting the the average usefulness of a model. In my experience RMSE is often lower than MAE, so it can artificially make it look like your model is doing better you think.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.