# Calculating per cent improvement in test error

I’ve come across this statement regarding an improvement percentage and I don’t know if it’s okay. It says this:

The training error and test error using the neighborhood model are 0.34 and 0.54 respectively, compared to 0.51 and 0.64 using the baseline predictor. This represents a 16% improvement in RMSE for the test set.

How does it get that 16%?

I would do:

Initial : 0.64 Final :0. 54

final - initial = |0.54 - 0.64 |=0.10

0.10 / 0.64 = 0.1562

0.1562 x100 = 15.72%

Is that correct?

• You are comparing 0.54 (test error with your model) to 0.64 (test error with the baseline predictor), so your result is how much much the model improved the results in the test set. – Heikki Pulkkinen Jun 13 at 9:56
• I get that! It was confusing of how it calculated that 16%. I wanted to be sure I was doing the right calculation. – Delan Jun 13 at 9:58