On which step should use SMOTE technique for over sampling?

Question

I want to use SMOTE technique for over sampling but I don't know on which step on pre-processing I should use it.

My preprocessing steps are:

• Missing values

• Removing Outliers

• Smoothing Data

Should I use SMOTE before all of these steps or its better to use it after these steps?

Answer 1

If you are using python, you can't use SMOTE in the presence of null values.

In this case:

1. Remove Outliers
2. Smooth Data
3. Impute null values (there are some smart options for that in R: using random forests to impute)
4. SMOTE

Removing outliers first let you do better smoothing and imputing.