0
$\begingroup$

I am working with a relatively small (~1000 samples) dataset that has some very messy text data (i.e lots of things missing, no real structure, etc.). I am trying to preprocess it and just going through each data sample and preprocessing it by hand is a viable option. I think that writing an algorithm to automate the process would actually end up being more tedious and take longer than just getting my hands dirty with the data.

So, my question is: is it common for data scientists/academics/professionals to preprocess a relatively small dataset by hand if it is the best and quickest option, or is anything done without automation looked down upon?

$\endgroup$
1
$\begingroup$

If you have a really small dataset like a few 100 samples it's okay to do preporcessing via hand. But since you have a thousand samples it's better to automate the process. You can use the na_values attribute in pandas to fill in the "???" or "??" value with nan. Then for each column replace the Nan values using some statistical measure (in your case the most occuring word..etc). But if youre data-wrangling is unworkable via an automation process or takes too much time, then you should proceed manually. If it's about meeting a deadline a deadline the manual way works best for really messy data.

$\endgroup$

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.