0
$\begingroup$

I'm working on a project about predicting kickstarter project success(classification) and my dataset has many columns that could be used as features such as : state_changed_at, launched_at, created_at.

Now the dataset has these features on unix timestamps.

Do I need to convert dates to some other format?

Can date data be used as a feature ?

If so how do I handle them ?

Do I keep them as a unix timestamp and try to scale/normalize them ?

$\endgroup$
1
$\begingroup$

Can date data be used as a feature?

Yes.

If so how do I handle them ?

Think about your problem. Why should the date be a reasonable indicator for the success of a startup? Answering this question tells you also in which way you need to transform it.

Most often, when I use some date information for models, I do the following:

  • Day of the week: integer/one-hot encoding for Monday, Tuesday, ..., Sunday
  • Month: integer/one-hot encoding for January, February, ..., December
  • Hour of the day: integer/one-hot encoding for 0, ..., 23
  • Seconds/minutes/hours/days since XY: Usually normalized or at least scaled in some way
| improve this answer | |
$\endgroup$

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.