Pandas 操作将 dtype: timedelta64[ns] 更改为 dtype: object

Question

I am calculating time differences in a pandas data frame so that I can then sum these time differences. Here is my full code for doing that:

temp['ACCESS_TIME'] = pd.to_datetime(temp['ACCESS_TIME'])
temp = temp.sort_values(['ID','ACCESS_TIME'])

temp['TIME_DIFFERENCE'] =  temp.groupby(['ID'])['ACCESS_TIME'].diff()

oneMin = temp.copy()

oneMin.loc[(oneMin.TIME_DIFFERENCE > '00:01:00'), 'TIME_DIFFERENCE'] = '00:01:00'
oneMin['DURATION'] = oneMin.groupby(['ID'])['TIME_DIFFERENCE'].transform('sum')

After the third line, a simple "temp['TIME_DIFFERENCE']" shows dtype: timedelta64[ns]. But after the 5th line, printing "oneMin['TIME_DIFFERENCE']" shows dtype: object. This leads to an error when attempting line 6, which states: "unsupported operand type(s) for +: 'int' and 'Timedelta'". I don't really know what is going on.

The issue is that I want to cap time differences at 1 minute. Before I was just removing time differences of 1 minutes with this line:

oneMin = oneMin[oneMin['TIME_DIFFERENCE'] < '00:01:00']

this works completely fine. But I do not want to remove the values over a minute, I want to cap them at one minute.

Answer 1

The problem is that you are setting some of the values of the column 'TIME_DIFFERENCE' to the string '00:01:00'. What you probably want to do is replace the fifth line in your code with

oneMin.loc[(oneMin.TIME_DIFFERENCE > '00:01:00'), 'TIME_DIFFERENCE'] = pd.Timedelta('00:01:00')