CODEWITHSUNDEEP
pythonpandasdatetimetime
JLuu
JLuu

Reputation: 373

unable to declare time object in python

I'm a bit new to this and I'm working with datetime data in python. Two questions:

I have a time column that is associated with events, but I'm having difficulty declaring it as time using datetime.time. I have a time column that is formatted like this:

0       11:17:43
1       06:00:00
2       06:30:35
3       02:00:00
4       23:00:00
5       13:20:49
6       19:30:00

and am attempting to declare it as a time object

data['timeobject'] = datetime.time(data['start_time'], axis = 1)

But am getting this error message:

TypeError: cannot convert the series to class 'int'

Also, I'd like to take this time object and use as a dependent variable using scikit-learn classification or regression.

How do I declare it a time object, and would there be any issues running it through scikit-learn models to predict when an event might happen?

Thank you!

Upvotes: 0

Views: 68

Answers (2)

jpp
jpp

Reputation: 164793

datetime.time does not work in a vectorised fashion. Pandas top-level pd.to_timedelta does, and it accepts a wide range of formats, including strings in the format you have supplied. Given a dataframe with column 'td':

df['td'] = pd.to_timedelta(df['td'])

print(df)

0   11:17:43
1   06:00:00
2   06:30:35
3   02:00:00
4   23:00:00
5   13:20:49
6   19:30:00
Name: td, dtype: timedelta64[ns]

Underlying the resultant series is an integer array via np.timedelta64. You should expect this to work well with the scikit-learn framework.

Upvotes: 1

Uri Goren
Uri Goren

Reputation: 13700

You should use pandas.to_datetime, and not the standard library datetime

data['timeobject'] = pd.to_datetime(data['start_time'], format='%h:%m:%s')

Upvotes: 1

Related Questions