Bigger mistakes, bigger penalty
All errors are wrong, but not all are equally bad. Sometimes large prediction errors are disproportionately more harmful than small errors.
Bigger mistakes, bigger penalty - that’s one of the features of the root mean squared error or RMSE. It squares large errors, which punishes these outliers more harshly than smaller errors.
RMSE can be calculated using the following formula, where the \(i\) th squared_diff is the square of the \(i\) th error.
$$RMSE = \sqrt{\frac{1}{n} \cdot \sum_{i=1} ^n i\text{th squared_diff}}$$
In this exercise, you will compute the RMSE of your predictions.
Available in your workspace is the result of the last exercise, test_enriched, the test data with a new column .pred, the model's out-of-sample predictions.
This exercise is part of the course
Machine Learning with Tree-Based Models in R
Exercise instructions
- Calculate the component-wise differences of the predictions and the final grades, square them, and save as
squared_diffs. - Use the formula above to calculate the RMSE and save it as
rmse_manual. - Use the
rmse()function to calculate the error and save asrmse_auto. - Print
rmse_manualandrmse_autoto verify that they are the same.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Calculate the squared differences
squared_diffs <- (___ - ___)^___
# Compute the RMSE using the formula
rmse_manual <- ___(1 / ___ * ___)
# Compute the RMSE using a function
rmse_auto <- ___(___,
___,
___)
# Print both errors
___
___