Skip to main content
 首页 » 编程设计

python之如何准确计算每天 DataFrame 中 NaN 的百分比变化

2025年01月19日9tintown

我想计算此 DataFrame (frame_) 的每日百分比变化:

import pandas as pd 
import numpy as np 
data_ = { 
    'A':[1,np.NaN,2,1,1,2], 
    'B':[1,2,3,1,np.NaN,1], 
    'C':[1,2,np.NaN,1,1,2], 
       } 
dates_ = [ 
    '06/01/2018','05/01/2018','04/01/2018','03/01/2018','02/01/2018', '01/01/2018' 
    ] 
 
frame_ = pd.DataFrame(data_, index=dates_, columns=['A','B','C']) 

问题是我用这种方法得到了一个 DataFrame:

returns_ = frame_.pct_change(periods=1, fill_method='pad') 
 
dates,A,B,C 
06/01/2018,,, 
05/01/2018,,1.0,1.0 
04/01/2018,1.0,0.5, 
03/01/2018,-0.5,-0.6666666666666667,-0.5 
02/01/2018,0.0,,0.0 
01/01/2018,1.0,0.0,1.0 

这不是我要找的。而且 dropna() 方法也没有给我想要的结果。我想为没有值(value)或 NaN 的那一天计算一个有值(value)和 NaN 的每一天的值(value)。例如,在 A 列上:我希望看到的百分比变化

dates,A 
06/01/2018,1 
05/01/2018, 
04/01/2018,1.0 
03/01/2018,-0.5 
02/01/2018,0.0 
01/01/2018,1.0 

提前致谢

请您参考如下方法:

这是一种方法,有点暴力。

import pandas as pd 
import numpy as np 
data_ = { 
    'A':[1,np.NaN,2,1,1,2], 
    'B':[1,2,3,1,np.NaN,1], 
    'C':[1,2,np.NaN,1,1,2], 
       } 
dates_ = [ 
    '06/01/2018','05/01/2018','04/01/2018','03/01/2018','02/01/2018', '01/01/2018' 
    ] 
 
frame_ = pd.DataFrame(data_, index=dates_, columns=['A','B','C']) 
frame_ = pd.concat([frame_, pd.DataFrame(columns=['dA', 'dB', 'dC'])]) 
 
for col in ['A', 'B', 'C']: 
    frame_['d'+col] = frame_[col].pct_change() 
    frame_.loc[pd.notnull(frame_[col]) & pd.isnull(frame_['d'+col]), 'd'+col] = frame_[col] 
 
#               A    B    C   dA        dB   dC 
# 06/01/2018  1.0  1.0  1.0  1.0  1.000000  1.0 
# 05/01/2018  NaN  2.0  2.0  NaN  1.000000  1.0 
# 04/01/2018  2.0  3.0  NaN  1.0  0.500000  NaN 
# 03/01/2018  1.0  1.0  1.0 -0.5 -0.666667 -0.5 
# 02/01/2018  1.0  NaN  1.0  0.0       NaN  0.0 
# 01/01/2018  2.0  1.0  2.0  1.0  0.000000  1.0