1
$\begingroup$

I am trying a binary classification on a Time Series data using XGBOOST and have sales variables with different currencies for different customers. Will different currencies (basically variable with different range) have a significant impact on the model? The two solutions I could think of is:

  1. Conversion for all rows for sales variable in a single currency.
  2. Adding a region of sales/currency as a categorical variable.

Will the second solution work or I should stick with the first one.

$\endgroup$
2
$\begingroup$

The first solution is valid. You need to keep the same unit for all rows per a given feature. But for currencies, you should be careful about the rate of conversion which might be dependent on the date of the data collection.

For the second solution, I am not quite sure about your end application but you may use it if you are trying to incorporate the effect of currency type in your model prediction. But if you are proposing it as alternative to the first solution, your model will less likely to infer the conversion rate among the various currency types.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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