0
$\begingroup$

I have some doubts on when to perform eda and the preprocessing the dataset should be put through.

  1. Should EDA be performed before or after balancing the dataset?

  2. Should one hot encoding and label encoding be applied to factor variables before EDA?

I guess what I'm asking is: should eda be performed on the dataset that is ready to be fed to my ml model?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
4
$\begingroup$

Exploratory data analysis is about exploring the data to learn more about it and understand it better. It covers a wide variety of techniques like calculating statistics, plotting, querying, eyeballing the data, etc. Before doing exploratory data analysis you don't know if your data is balanced or not (by the way, unbalanced data is a tricky topic) because you didn't look at the data yet! You also don't know if you need to apply one-hot-encoding, because you don't know what the data is, as you didn't look at it yet! The moment you look at the results of your first SQL query or looked at the content of the CSV file, you already started the exploratory data analysis.

Improve this answer
$\endgroup$
2
  • $\begingroup$ So If I get this right, EDA is literally the first thing that takes place and It shouldn't be affected by under/over balancing or any type of encoding. It's a pure analysis of what I got. $\endgroup$
    – IDK
    Commented Aug 28, 2022 at 13:38
  • $\begingroup$ @IDK if you're going for an exploratory trip to the jungle, your trip will be affected by what the jungle is, but your point is to explore it and find out what it is. $\endgroup$
    – Tim
    Commented Aug 28, 2022 at 14:50

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.