I have a data set where certain features are time stamps for certain important events like time stamp of when a user logged in etc. How should I deal with null values in such cases. Should one just replace null values with current or most recent time stamp. Are there any standard techniques followed ?
1
-
$\begingroup$ You should edit your question to precise the context. The way you deal with missing value highly depends of the whole process (data acquisition, further treatments, goal of the study....) $\endgroup$– Manu HCommented Sep 6, 2016 at 12:08
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
Leave those time stamps null in order to indicate that you do not possess the time of those events. Everything else depends on the logic of your application.
$\begingroup$
$\endgroup$
1
It is not clear from your question what your use case is exactly. But for cases like what you mentioned, it is usually more helpful if you generate time series out of each feature. For example, if you have user's login and logout time, you can generate a time series where it is 1 when the user is logged in and 0 when the user is logged out. But again, tell us more about your use case first.
-
$\begingroup$ In my case I have a data set and in that few of the features happen to be time stamps, ex: last logged in timestamp, account created time stamp etc. Should I replace the null values with like an average of the all non-null time stamps in that column. $\endgroup$ Commented Sep 6, 2016 at 10:51