Tukukkk

I have a dataset of column name DateTime having dtype object.

df['DateTime'] = pd.to_datetime(df['DateTime'])

I have used the above code to convert to datetime format then did a split in the column to have Date and Time separately

df['date'] = df['DateTime'].dt.date

df['time'] = df['DateTime'].dt.time

but after the split the format changes to object type and while converting it to datetime it showing error for the time column name as: TypeError: is not convertible to datetime

How to convert it to datetime format the time column

You can use combine in list comprehension with zip:

df = pd.DataFrame({'DateTime': ['2011-01-01 12:48:20', '2014-01-01 12:30:45']})

df['DateTime'] = pd.to_datetime(df['DateTime'])



df['date'] = df['DateTime'].dt.date

df['time'] = df['DateTime'].dt.time



import datetime

df['new'] = [datetime.datetime.combine(a, b) for a, b in zip(df['date'], df['time'])]

print (df)



             DateTime        date      time                 new

0 2011-01-01 12:48:20  2011-01-01  12:48:20 2011-01-01 12:48:20

1 2014-01-01 12:30:45  2014-01-01  12:30:45 2014-01-01 12:30:45

Or convert to strings, join together and convert again:

df['new'] = pd.to_datetime(df['date'].astype(str) + ' ' +df['time'].astype(str))

print (df)

             DateTime        date      time                 new

0 2011-01-01 12:48:20  2011-01-01  12:48:20 2011-01-01 12:48:20

1 2014-01-01 12:30:45  2014-01-01  12:30:45 2014-01-01 12:30:45

But if use floor for remove times with converting times to timedeltas then use + only:

df['date'] = df['DateTime'].dt.floor('d')

df['time'] = pd.to_timedelta(df['DateTime'].dt.strftime('%H:%M:%S'))



df['new'] = df['date'] + df['time']

print (df)



             DateTime       date     time                 new

0 2011-01-01 12:48:20 2011-01-01 12:48:20 2011-01-01 12:48:20

1 2014-01-01 12:30:45 2014-01-01 12:30:45 2014-01-01 12:30:45

How to convert it back to datetime format the time column

There appears to be a misunderstanding. Pandas datetime series must include date and time components. This is non-negotiable. You can simply use pd.to_datetime without specifying a date and use the default 1900-01-01 date:

# date from jezrael



print(pd.to_datetime(df['time'], format='%H:%M:%S'))



0   1900-01-01 12:48:20

1   1900-01-01 12:30:45

Name: time, dtype: datetime64[ns]

Or use another date component, for example today's date:

today = pd.Timestamp('today').strftime('%Y-%m-%d')

print(pd.to_datetime(today + ' '  + df['time'].astype(str)))



0   2018-11-25 12:48:20

1   2018-11-25 12:30:45

Name: time, dtype: datetime64[ns]

Or recombine from your date and time series:

print(pd.to_datetime(df['date'].astype(str) + ' ' + df['time'].astype(str)))



0   2011-01-01 12:48:20

1   2014-01-01 12:30:45

dtype: datetime64[ns]