0
$\begingroup$

I'm new to ML and researching data prep, more specifically feature normalisation. My question is whether it's a good idea to normalise data when its range may change over time?

For example, if I'm trying to predict stock prices in my train dataset, the prices range from 100 to 200 and later (unseen data) they could reach 300.

Should I normalise the training data?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
2
$\begingroup$

whether it's a good idea to normalise data when its range may change over time?

Short answer : Yes

Long answer : It is generally a good idea to scale/normalise your data in machine learning.

However, I am not sure what you mean when you say :

when its range may change over time?

I would assume that you are asking if the range of dependent variable(y) changes over time. Note that you don’t necessarily scale the dependent variable(y), just the independent variables.So, In this case you won’t have to anything additionally.

However, there is one boundary case that you might need to consider : If the change in range of dependent variable (y) is because of some change in distribution of data, you might need to go back, do the necessary data preparation and hence the normalisation and train your model again.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.