从Python现有序列中生成扩展的时间序列数据

Question

Well, I searched for the to this solution for long but I can't find it, even though I believe it will be something pretty easy to do. I have a time series, in 1 hr intervals for one year. What I want to do is to create fake data for next years, by tinkering with my original data just a bit. For example, if my original data looks like that

Date standard   Estimated production 

1/1/2016 7:00   0,0  
1/1/2016 8:00   0,0  
1/1/2016 9:00   16,3  
1/1/2016 10:00  29,4   
1/1/2016 11:00  40,6  
1/1/2016 12:00  33,9

(it continues like that until the end of the year), I would like to create fake data that for each respective date, is similar.

Date standard   Estimated production 

1/1/2017 7:00   0,01  
1/1/2017 8:00   0,03  
1/1/2017 9:00   16,1 
1/1/2017 10:00  29,3  
1/1/2017 11:00  40,8  
1/1/2017 12:00  33,1

The above changes are of course totally random, the production should be increased or decreased by a number within a set limit. Thank you in advance!

Answer 1

You could use DateOffset to shift the index by one year (and then you can modify the values as you want).

To generate noise, you could look at numpy random utils.

import numpy as np
# Same values as 2016, but dates shifted by 1 year (2017)
fake_data = df.loc['2016'].copy()
fake_data.index = fake_data.index + pd.DateOffset(years=1)

# Add gaussian noise, with same standard deviation of production
noise = np.random.randn(len(fake_data)) * fake_data['production'].std()
fake_data['production'] = fake_data['production'] + noise

new_data = pd.concat([data, new_data], axis=0)