Converting ints to timedelta with pandas

If you are getting all 00:00:00.000000 values, then your air_time values might be strings. (You can check the data type of the air_time column by inspecting data.info(). If the dtype says object then the values are Python objects (such as strs) instead of a NumPy integer data type. You can then confirm they are strings by inspecting set(map(type, data['air_time'])).)

If they are strings, you can convert them to ints first by using:

data['air_time'] = data['air_time'].astype(int)

If 137 means 137 minutes then use

data['air_time'] = pd.to_timedelta(data['air_time'], unit='m', errors='coerce')

If, on the other hand, 137 means 1 hour and 37 minutes, then use

data['air_time'] = pd.to_timedelta(
    (data['air_time']//100)*60 + (data['air_time'] % 100), unit='m', 
    errors='coerce')

The unit='m' argument tells pd.to_timedelta to interpret the values as minutes.

For example,

import pandas as pd

data = pd.DataFrame({'air_time':['137','137','126','92','352']})
data['air_time'] = data['air_time'].astype(int)
data['air_time'] = pd.to_timedelta(data['air_time'], unit='m', errors='coerce')

yields

  air_time
0 02:17:00
1 02:17:00
2 02:06:00
3 01:32:00
4 05:52:00

Note that pd.to_timedelta can also accepts strings as input if the strings contain the desired units. For example,

import pandas as pd

data = pd.DataFrame({'air_time':['137','137','126','92','352']})
data['air_time'] = data['air_time'] + ' minutes'
#       air_time
# 0  137 minutes
# 1  137 minutes
# 2  126 minutes
# 3   92 minutes
# 4  352 minutes

data['air_time'] = pd.to_timedelta(data['air_time'], errors='coerce')

yields the same result.