Datetime columns are really hard to be identified by pandas and they are assigned with dtype="object".
Since pytrousse package wants to avoid the manual analysis of each column with dtype="object", it could be useful to include a property that returns the possible columns containing datetime values.
My proposal to do so it to try to convert to datetime values the columns with dtype="object" (by using pd.to_datetime() function) and a check can be performed on how many values are lost on those columns. This information can be shared with the user that can decide which columns should really be converted or not.
Datetime columns are really hard to be identified by
pandasand they are assigned withdtype="object". Since pytrousse package wants to avoid the manual analysis of each column withdtype="object", it could be useful to include a property that returns the possible columns containing datetime values.My proposal to do so it to try to convert to datetime values the columns with
dtype="object"(by usingpd.to_datetime()function) and a check can be performed on how many values are lost on those columns. This information can be shared with the user that can decide which columns should really be converted or not.