将数据框与其他数据框合并并根据特定条件计算分组百分比

Posted 2023-03-11

【中文标题】将数据框与其他数据框合并并根据特定条件计算分组百分比

【英文标题】：Merge data frame with other and calculate groupby percentage based on the specific condition

【发布时间】：2021-09-10 13:06:03

【问题描述】：

我有两个数据框，如下图

df1:

Sports     Expected_%
Cricket    70
Football   20
Tennis     10

df2:

Region    Sports     Count    Percentage     
North     Cricket    800      75                              
North     Football   50       5            
North     Tennis     150      20           
South     Cricket    1300     65           
South     Football   550      27.5         
South     Tennis     150      7.5  

    

预期输出：

Region    Sports     Count    Percentage   Expected_%     Expected_count    
North     Cricket    800      75           70             700
North     Football   50       5            20             200
North     Tennis     150      20           10             100
South     Cricket    1300     65           70             1400
South     Football   550      27.5         20             400
South     Tennis     150      7.5          10             200

解释：

Expected_% for Cricket = 70

Total Count for North = 1000

Expected_Count for North = 1000*70/100 = 700

【参考方案1】：

将DataFrame.merge 与左连接用于新列，然后将GroupBy.transform 与sum 用于新Series，乘以新列并除以100：

df = df2.merge(df1, on='Sports', how='left')
summed = df.groupby('Region')['Count'].transform('sum')
df['Expected_count'] = summed.mul(df['Expected_%']).div(100)
print (df)
  Region    Sports  Count  Percentage  Expected_%  Expected_count
0  North   Cricket    800        75.0          70           700.0
1  North  Football     50         5.0          20           200.0
2  North    Tennis    150        20.0          10           100.0
3  South   Cricket   1300        65.0          70          1400.0
4  South  Football    550        27.5          20           400.0
5  South    Tennis    150         7.5          10           200.0

或使用Series.map 新建列：

df2['Expected_%']= df2['Sports'].map(df1.set_index('Sports')['Expected_%'])
summed = df2.groupby('Region')['Count'].transform('sum')
df2['Expected_count'] = summed.mul(df2['Expected_%']).div(100)
print (df2)
  Region    Sports  Count  Percentage  Expected_%  Expected_count
0  North   Cricket    800        75.0          70           700.0
1  North  Football     50         5.0          20           200.0
2  North    Tennis    150        20.0          10           100.0
3  South   Cricket   1300        65.0          70          1400.0
4  South  Football    550        27.5          20           400.0
5  South    Tennis    150         7.5          10           200.0

【参考方案2】：

另一种方式：

map_dict = dict(df1.values)
df2['Percentage'] = df2.groupby('Region').apply(lambda x: (x['Count'].sum() * x['Sports'].map(map_dict))).div(100).values

【讨论】：

性能...避免这种情况，因为速度慢。

@jezrael Ohh!!.. 我还没有检查过这个性能！！

是的，取决于数据，我想慢 10 倍，但也许更多。

简单的一般规则 - 原生 pandas 函数很快，自定义函数不是。如果每组调用一个有点复杂的函数，那么性能会降低，因为groupby.apply，因为复杂的函数，以及因为调用 N 次（组数）s

@jezrael 有道理！！