基于两个分类列的累积计数

Question

我想为表中的每个记录基于两个分类列进行累计计数。

在下表中，我想获取 cum_count 列，该列是根据 industry 和 deal_status 列计算的。这个想法是，针对每条记录，计算同一行业的上一个获胜交易的数量。

例如，该表的最后一条记录的 cum_count = 3，因为只有3个 deal_status =赢得了 industry = x的交易以前见过。

Pandas' GroupBy.cumcount function仅针对单个变量执行此操作...

如何针对我所描述的情况来做到这一点？

pd.DataFrame({'time': [1, 2, 3, 4, 5, 6, 7],
              'company' : ["ciaA", "ciaB", "ciaA", "ciaC", "ciaA", "ciaD", "ciaE"],
              'industry' : ["x", "y", "x", "x", "x", "y", "x"],
              'deal_status' : ["won", "lost", "won", "won", "lost", "won", "lost"],
              'cum_count' : [0, 0, 1, 2, 3, 0, 3]})


time    company    industry     deal_status     cum_count
 1       ciaA         x             won             0
 2       ciaB         y            lost             0
 3       ciaA         x             won             1
 4       ciaC         x             won             2
 5       ciaA         x            lost             3
 6       ciaD         y             won             0
 7       ciaE         x            lost             3

Answer 1

创建一个助手列，您将使用该累加器的总和。需要在每个组中移动，因为您的计数仅包括上一个获胜值：

df['to_sum'] = (df.deal_status == 'won').astype(int)
df['cum_count'] = (df.groupby('industry')
                    .apply(lambda x: x.to_sum.shift(1).cumsum()).fillna(0)
                    .reset_index(0, drop=True))

输出df：

   time company industry deal_status  to_sum  cum_count
0     1    ciaA        x         won       1        0.0
1     2    ciaB        y        lost       0        0.0
2     3    ciaA        x         won       1        1.0
3     4    ciaC        x         won       1        2.0
4     5    ciaA        x        lost       0        3.0
5     6    ciaD        y         won       1        0.0
6     7    ciaE        x        lost       0        3.0