熊猫:使用滚动功能检查NaN

时间:2019-03-27 17:18:17

标签: pandas apply nan rolling-computation

我有一个带有变量“ A”的数据框,我想创建一个滚动的Nan检查器,以便如果所有3(秒)个单元(当前单元格和前两个单元格)的新变量“ rolling_nan” = 1 )是NaN,否则“ rolling_nan” = 0。

我正在应用一个函数,因为.rolling熊猫函数不支持isna()。但是我得到以下。另外,我不确定如何在NaN检查器中包含相同的行值。

import pandas as pd
import numpy as np

idx = pd.date_range('2018-01-01', periods=10, freq='S')
df = pd.DataFrame({"A":[1,2,3,np.nan,np.nan,np.nan,6,7,8,9]}, index = idx)
df

def isna_func(x):
    return 1 if pd.isna(x).all() == True else 0
df['rolling_nan'] = df['A'].rolling(3).apply(isna_func)
df

                    A   rolling_nan
2018-01-01 00:00:00 1.0 NaN
2018-01-01 00:00:01 2.0 NaN
2018-01-01 00:00:02 3.0 0.0
2018-01-01 00:00:03 NaN NaN
2018-01-01 00:00:04 NaN NaN
2018-01-01 00:00:05 NaN NaN
2018-01-01 00:00:06 6.0 NaN
2018-01-01 00:00:07 7.0 NaN
2018-01-01 00:00:08 8.0 0.0
2018-01-01 00:00:09 9.0 0.0

在上面的示例中,rolling_nan仅应在时间戳2018-01-01 00:00:05等于1,否则应等于0。

1 个答案:

答案 0 :(得分:1)

您可以用不同的方式标记所有notna并找到max

df.A.notna().rolling(3).max()==0
Out[316]: 
2018-01-01 00:00:00    False
2018-01-01 00:00:01    False
2018-01-01 00:00:02    False
2018-01-01 00:00:03    False
2018-01-01 00:00:04    False
2018-01-01 00:00:05     True
2018-01-01 00:00:06    False
2018-01-01 00:00:07    False
2018-01-01 00:00:08    False
2018-01-01 00:00:09    False
Freq: S, Name: A, dtype: bool

重新分配

df['rollingnan']=(df.A.notna().rolling(3).max()==0).astype(int)
df
Out[320]: 
                       A  rollingnan
2018-01-01 00:00:00  1.0           0
2018-01-01 00:00:01  2.0           0
2018-01-01 00:00:02  3.0           0
2018-01-01 00:00:03  NaN           0
2018-01-01 00:00:04  NaN           0
2018-01-01 00:00:05  NaN           1
2018-01-01 00:00:06  6.0           0
2018-01-01 00:00:07  7.0           0
2018-01-01 00:00:08  8.0           0
2018-01-01 00:00:09  9.0           0

或者根据您自己的想法使用all

df['A'].isna().rolling(3).apply(lambda x : x.all(),raw=True)
Out[323]: 
2018-01-01 00:00:00    NaN
2018-01-01 00:00:01    NaN
2018-01-01 00:00:02    0.0
2018-01-01 00:00:03    0.0
2018-01-01 00:00:04    0.0
2018-01-01 00:00:05    1.0
2018-01-01 00:00:06    0.0
2018-01-01 00:00:07    0.0
2018-01-01 00:00:08    0.0
2018-01-01 00:00:09    0.0
Freq: S, Name: A, dtype: float64