熊猫时间序列重采样+线性调整值

Question

Using python and pandas, how do I resample a time series to even 5-min intervals (offset=zero min from whole hours) while also adjusting the values linearly?

Hence, I want to turn this:

         value
00:01    2
00:05    10
00:11    22
00:14    28

into this:

         value
00:00    0
00:05    10
00:10    20
00:15    30

Please note how the "value"-column was adjusted.

• For simplicity, I have chosen the values to be exactly 2 * number of minutes.
• In real life, however, the values are not that perfect. Sometimes there will exist more than one value between two even 5-min intervals and sometimes more than one 5-min interval between two "real" values, so when resampling I need to, for each even 5-min interval, find the "real" values before and after that even 5-min interval, and calculate a linearly interpolated value from them.

PS.

There is a lot of information about this everywhere on the inte.net, but I still wasn't able to find a function (sum, max, mean, etc, or write my own functino) that could accompish what I wanted to do.

Answer 1

I have reconsidered the code because the requirement was omitted from the comments. Create a new data frame by combining the original data frame with a data frame that is extended to one minute. I linearly interpolated the new data frame and extracted the results in 5-minute increments. This is my understanding of the process. If I'm wrong, please give me another answer.

import pandas as pd
import numpy as np
import io

data = '''
time value
00:01 2
00:05 10
00:11 22
00:14 28
00:18 39
'''
df = pd.read_csv(io.StringIO(data), sep='\s+')
df['time'] = pd.to_datetime(df['time'], format='%H:%M')
time_rng = pd.date_range(df['time'][0], df['time'][4], freq='1min')
df2 = pd.DataFrame({'time':time_rng})
df2 = df2.merge(df, on='time', how='outer')
df2 = df2.set_index('time').interpolate('time')
df2.asfreq('5min')
    value
time    
1900-01-01 00:01:00 2.0
1900-01-01 00:06:00 12.0
1900-01-01 00:11:00 22.0
1900-01-01 00:16:00 33.5

Answer 2

You can use datetime and time module to get the sequence of time intervals. Then use pandas to convert the dictionary into a dataframe. Here's the code to do that.

import time, datetime
import pandas as pd

#set the dictionary as time and value
data = {'Time':[],'Value':[]}

#set a to 00:00 (HH:MM) 
a = datetime.datetime(1,1,1,0,0,0)

#loop through the code to create 60 mins. You can increase loop if you want more values
#skip by 5 to get your 5 minute interval

for i in range (0,61,5):
    # add the time and value into the dictionary
    data['Time'].append(a.strftime('%H:%M'))
    data['Value'].append(i*2)

    #add 5 minutes to your date-time variable

    a += datetime.timedelta(minutes=5)

#now that you have all the values in dictionary 'data', convert to DataFrame
df = pd.DataFrame.from_dict(data)

#print the dataframe
print (df)

#for your reference, I also printed the dictionary
print (data)

The dictionary will look as follows:

{'Time': ['00:00', '00:05', '00:10', '00:15', '00:20', '00:25', '00:30', '00:35', '00:40', '00:45', '00:50', '00:55', '01:00'], 'Value': [0, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120]}

The dataframe will look as follows:

     Time  Value
0   00:00      0
1   00:05     10
2   00:10     20
3   00:15     30
4   00:20     40
5   00:25     50
6   00:30     60
7   00:35     70
8   00:40     80
9   00:45     90
10  00:50    100
11  00:55    110
12  01:00    120