05-18-2017 10:52 PM - edited 05-18-2017 11:00 PM
I got an issue when using Extract Macro with datetime values. The datetime value i want to extract is "Dec 03, 2016 9:04:51 AM", but after using Extract Macro, the value extracted is "03/12/2016 09:04:51 AM" (I used Log operator to see this value), which mean "Mar 12, 2016 9:04:51 AM" (day becomes month and month becomes day). And then when I used Filter Examples to keep only records after "Dec 03, 2016 9:04:51 AM", i got all records after "Mar 12, 2016 9:04:51 AM", which is not what i want. Do you have any idea to solve this? Thank you very much.
I attached my process here.
Solved! Go to Solution.
05-19-2017 07:28 AM
05-19-2017 07:32 AM
I'm not saying this isn't a problem, but it sounds like a date format issue--the difference described is between the American vs the European way of writing dates (3/12/16 is March 12th to an American and Dec 3rd to a European). The date format could be modified in RapidMiner using Date to Nominal and then Nominal to Date using a different format. There may be a more efficient way of solving this though.
05-21-2017 09:55 PM - edited 05-21-2017 10:15 PM
The datetime i used is under RapidMiner format. I attached the input data in this message.
The problem here i think is the inconsistence between "Extract Macro" (EM) and "Filter Examples" (FE). FE uses the datetime value from EM but it reads the value in a different format (as what Brian explained above).
05-22-2017 05:42 AM
I recommend using Generate Attributes to create a numerical Attribute representing the date value in milliseconds since Jan, 1st 1970 (see screenshot).
You can then apply Extract Macro and Filter Examples on this attribute.
05-22-2017 07:07 AM
Thanks Edin, this should work.
05-22-2017 10:54 AM
You can do a similar extraction directly with the "Date to Numeric" operator, and choose units other than milliseconds as well (if you'd rather think about this in a unit that is not as large in absolute terms or that is more meaningful to you like days or weeks).