基于日期的多条件计数器

Question

我有这个数据框

df:
    entrance   leaving        counter
1   2012-07-01  NaT             NaN
2   2013-03-15  NaT             NaN
3   2013-03-15  2013-04-15      NaN
4   2014-06-01  NaT             NaN
5   2014-06-01  NaT             NaN

我想要考虑两列日期的计数器，并在 entrance 日期递增，在有 leaving 日期时递减。此外，下面的 date 列也应增加一个月。所需的输出应为：

df_new:
date      counter
2012-07     1
2012-08     1              
  ...      ...             
2013-03     2
  ...      ...
2014-06     4

我在这行中根据 entrance 递增，但我无法使用 np.where() 来递减 `df.entrance.notnull()'。

df.groupby([df['entrance'].dt.to_period("M")]).entrance.count().cumsum()

Answer 1

我认为您的问题未指定。计数器不能共享原始 DF 的索引。以下是原因的示例：

    # Lets assume this is the DF:
    entrance   leaving        counter
1   2012-07-01  NaT             1
2   2013-03-15  NaT             2
3   2013-03-15  2013-06-15      2 ?
4   2013-06-01  NaT             3 or 4? Depends if you count the exit in prev row or not

不管怎样，解决方法如下：

# Load Data
s = '''     entrance   leaving        counter
1   2012-07-01  NaT             NaN
2   2013-03-15  NaT             NaN
3   2013-03-15  2013-04-15      NaN
4   2014-06-01  NaT             NaN
5   2014-06-01  NaT             NaN'''

df = pd.DataFrame.from_csv(io.StringIO(s), sep='\s+')
df['leaving']= pd.to_datetime(df['leaving'])
df['entrance']= pd.to_datetime(df['entrance'])

不会遵循原始索引的明确解决方案：

# Counter
counter = pd.Series(1, df['entrance'].dropna()).subtract(pd.Series(1, df['leaving'].dropna()), fill_value=0).cumsum()

# If you want it monthly
counter.resample('M').last().ffill()

保持原始索引但有些模糊的解决方案：

count_df = df.notna().cumsum()
df['counter'] = count_df['entrance'] - count_df['leaving']

基于日期的多条件计数器

multi-conditional counter based on dates

pandas

pandas-groupby

data-science

np