另一个解决方案：

df["Remaining Quatity"] = (g := df.groupby("Class").cumsum())["Received"] - g["Issued"]
print(df)

图纸：

Class  Received  Issued  Remaining Quatity
0    FD        10       0                 10
1    FD         0       2                  8
2    RM         5       0                  5
3    RM         0       3                  2
4    FD         0       2                  6
5    PM         5       0                  5
6    PM         1       0                  6
7    RM         1       0                  3
8    FD         4       0                 10

OR：使用.pipe：

df["Remaining Quatity"] = df.groupby("Class").cumsum().pipe(lambda g: g["Received"] - g["Issued"])

OR：使用.eval：

df["Remaining Quatity"] = df.groupby("Class").cumsum().eval("Received - Issued")

df['Remaining Quantity'] = df.groupby('Class').apply(
    lambda x: x['Received'].cumsum() - x['Issued'].cumsum()
    ).reset_index(level=0, drop=True)

输出量：

Class  Received  Issued  Remaining Quantity
0    FD        10       0                  10
1    FD         0       2                   8
2    RM         5       0                   5
3    RM         0       3                   2
4    FD         0       2                   6
5    PM         5       0                   5
6    PM         1       0                   6
7    RM         1       0                   3
8    FD         4       0                  10

另一种可能的解决方案：

df["Remaining Quatity"] = (
    df.eval("tmp=Received-Issued").groupby("Class")["tmp"].cumsum()
)

输出量：

print(df)

  Class  Received  Issued  Remaining Quatity
0    FD        10       0                 10
1    FD         0       2                  8
2    RM         5       0                  5
3    RM         0       3                  2
4    FD         0       2                  6
5    PM         5       0                  5
6    PM         1       0                  6
7    RM         1       0                  3
8    FD         4       0                 10

一种方法是使用.stack计算差值，然后沿着索引将值赋值。

df['Remaining Quality'] = df.assign(
            Issued=df['Issued'] * -1).set_index('Class',append=True)\
           .stack().groupby(level=1).cumsum().unstack(-1).droplevel(1,0)['Issued']

print(df)

  Class  Received  Issued  Remaining Quality
0    FD        10       0                 10
1    FD         0       2                  8
2    RM         5       0                  5
3    RM         0       3                  2
4    FD         0       2                  6
5    PM         5       0                  5
6    PM         1       0                  6
7    RM         1       0                  3
8    FD         4       0                 10