代码之家 › 专栏 › 技术社区 › SkyWalker

除月底外,最惯用的将所有值设置为NaN的方法是什么?

dataframe pandas python

SkyWalker · 技术社区 · 4 年前

我想学习最惯用的方法将数据帧的所有值设置为 NaN

# get the values corresponding to the last business day of each month
df_eofm = df.resample('BM').last()
# fill the original data frame with NaN's
df[:] = np.nan
# now try to set the last business days to the values we saved
df.update(df_eofm)
print(df)
print(df.dropna())

                        Col1                    Col2            Col3
Date                                                                
1963-12-31              57.5                     -28            0.89
1964-01-01               NaN                     NaN             NaN
1964-01-02               NaN                     NaN             NaN
1964-01-03               NaN                     NaN             NaN
1964-01-04               NaN                     NaN             NaN
...                      ...                     ...             ...
2020-03-11               NaN                     NaN             NaN
2020-03-12               NaN                     NaN             NaN
2020-03-13               NaN                     NaN             NaN
2020-03-14               NaN                     NaN             NaN
2020-03-15               NaN                     NaN             NaN

[20530 rows x 3 columns]
                        Col1                    Col2            Col3
Date                                                                
1963-12-31              57.5                     -28            0.89
1964-01-31                54                     106            0.65
1964-02-28              57.1                     126            0.68
1964-03-31              57.9                     266            0.73
1964-04-30              60.2                     144            0.72
...                      ...                     ...             ...
2019-10-31              47.8                     136            0.11
2019-11-29              48.3                     128            0.22
2019-12-31              48.1                     266            0.37
2020-01-31              47.2                     145           -0.08
2020-02-28              50.9                     225           -0.45

[675 rows x 3 columns]

1 回复 | 直到 4 年前

yatu Sayali Sonawane 4 年前

你可以用 is_month_end 并用生成的布尔序列索引数据帧:

df[~df.index.is_month_end] = np.nan

this answer 我们可以这样做:

def is_business_day(date):
    return bool(len(pd.bdate_range(date, date)))

last_bus = (df.index.to_frame()[0]    
              .map(is_business_day)
              .groupby(df.index.month)
              .transform(lambda x: x.last_valid_index()))
df[df.index==last_bus] = np.nan

推荐文章

July · 如何定义数字间隔,然后四舍五入

1 年前

Community wiki · 对象名称前的单下划线和双下划线的含义是什么?

1 年前