I have built a classification model using the following steps (and in the mentioned order) in Python -

  1. Data cleaning - Removing unwanted variables and separating Predictor variables from response variable
  2. Label Encoding
  3. Standardization( StandardScaler)
  4. Train Test Split
  5. Smote
  6. Model Building
  7. Model Testing using Test data

a) Now I have a new dataset, and I want to predict using the above built model. How do I do it ? Which of the above steps should I follow and which ones should I skip ? b) Also, is the arrangement of any of the steps aforementioned very, very wrong so that it needs to be changed ?


Welcome to Data Science on Stack Exchange,
You should apply the same steps 1-3 to the new dataset as you did to the original train and test datasets. Smote is a special case and couldn't be applied anyway as you don't have the response in the new dataset. In general, your train, test, validate, and any new datasets should go through the same data pre-processing. The new dataset will replace your Test dataset in your prediction.

There may be more steps you can add, but the current arrangement is fine.

  • $\begingroup$ Thanks for the info ! Also could you please answer part b of the question ? $\endgroup$
    – Saket123
    Jul 2 '20 at 13:49
  • $\begingroup$ sure, I updated my answer to include part b. $\endgroup$
    – Donald S
    Jul 2 '20 at 14:04

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.