0

Currently I am working on a project which uses Xgboost Regression.

Before putting data into model, I implemented Normalization, the accuracy significantly increased compared with without Normalization.

But I saw someone said that Xgboost doesn’t need Normalization .

Did someone meet the same issue as I met ?

Thanks in advance.

LordNR
  • 11
  • 1
  • 2
    Did you perform normalization before or after train test split? – Multivac May 05 '21 at 00:35
  • Feature scaling should have zero effect on the model (see e.g. https://datascience.stackexchange.com/q/77312/55122); could you elaborate on your process? – Ben Reiniger May 05 '21 at 02:32

1 Answers1

1

Xgboost is an ensemble algorithm based on decision trees, so doesn't need normalization. You can check this on Xgboost official github: Is Normalization necessary? and this post What are the implications of scaling the features to xgboost?

I'm new in this algorithm but I'm pretty sure of what I've written

simo954
  • 13
  • 5