从数据框中随机选择小时

Question

I have a hard time to randomly select rows from a dataframe. In general, choosing one row is not a problem using np.random.choice(data,size=1000) . I assume that replacement=True. However, I need to randomly select an hour and as output, recieve the 4 rows of each quarter.

The dataframe to choose from is the following (1132 rows):

data=
                     Price  Consume    Feed
StartTime                                  
2018-07-04 02:00:00  45.80    67.91   67.91
2018-07-04 02:15:00  45.80    51.05   51.05
2018-07-04 02:30:00  45.80    46.12   46.12
2018-07-04 02:45:00  45.80    46.86   46.86
2018-07-11 05:00:00  43.80    43.49   43.49
2018-07-11 05:15:00  43.80    50.71   50.71
2018-07-11 05:30:00  43.80    48.19   48.19
2018-07-11 05:45:00  43.80    40.02   40.02

My desired output is something like this:

Assuming the random generator has "selected" 2018-07-11 05:00:00 , the output would be

2018-07-11 05:00:00  43.80    43.49   43.49
2018-07-11 05:15:00  43.80    50.71   50.71
2018-07-11 05:30:00  43.80    48.19   48.19
2018-07-11 05:45:00  43.80    40.02   40.02

Depending on the number (N) of random samples, the length of the resulting dataframe should be 4xN.

Is it possible to randomly select an dayhour directly from the dataframe and repeat this 1000 times? I am afraid that using an extra dataframe to select an hour and then looking the corresponding values up in the original dataframe will be too time consuming. I am confident that this should be doable in Python, but I couldn`t find any tips on this.

Thanks for any help!

Answer 1

I think you compare values with DatetimeIndex.floor for remove minutes and seconds:

N = 1000
vals = pd.to_datetime(np.random.choice(df.index,size=N)).floor('H')
hours = df.index.floor('H')

for i in vals:
    print (df[hours == i])

EDIT: For join all small DataFrames use concat :

df1 = pd.concat([df[hours == i] for i in vals])

Or create array with np.concatenate of DatetimeIndex and select by loc :

idx = np.concatenate([df.index[hours == i] for i in vals])
df1 = df.loc[idx]

Answer 2

Sample once to get a random index, then find all matches to that date and hour:

random_idx = df.sample().index
df[(df.index.date == random_idx.date) & (list(df.index.hour) == random_idx.hour)]

Then to do it 1000 times:

for i in range(1000):
    random_idx = df.sample().index
    print(df[(df.index.date == random_idx.date) & (list(df.index.hour) == random_idx.hour)])