No split training and testing sets？ about gain HOT 2 CLOSED

I have another question, shouldn't we calculate RMSE within the original data range? Why are we calculating values within the normalized range. Looking forward to your reply, thank you.

from gain.

Thanks for your questions.

For the first issue, please see this link (#14).
For the second issue, yes, you can compute RMSE in the range of original data as well.

• In the paper, to properly compare across the datasets, we use the normalized data. However, it is definitely correct that we can compute RMSE in the original data range.
• One caveat is that if we use the original data range to compute RMSE, if one feature has very large range, it may be dominant for RMSE value.

from gain.