python - 如何使用 Groupby 函数查找数据框中的最高值

Question

我有以下数据集。我想查找研究中使用的每个应用在哪一年的季度中产生了最高的安装次数？

        Installs          CR     Month  Year             Category
0         10000    Everyone  January   2018       ART_AND_DESIGN
1        500000    Everyone  January   2018       ART_AND_DESIGN
2       5000000    Everyone   August   2018       ART_AND_DESIGN
3      50000000        Teen     June   2018       ART_AND_DESIGN
4        100000    Everyone     June   2018       ART_AND_DESIGN
        ...         ...       ...   ...                  ...
10836      5000    Everyone     July   2017               FAMILY
10837       100    Everyone     July   2018               FAMILY
10838      1000    Everyone  January   2017              MEDICAL
10839      1000  Mature 17+  January   2015  BOOKS_AND_REFERENCE
10840  10000000    Everyone     July   2018            LIFESTYLE

Answer 1

如果需要每季度和类别的最大值，请使用:

q = (pd.to_datetime(df['Month'] + df['Year'].astype(str), format='%B%Y')
       .dt.to_period('Q').rename('Quarter'))

df = df.groupby([q,'Category'])['Installs'].max().reset_index()
print (df)
  Quarter             Category  Installs
0  2015Q1  BOOKS_AND_REFERENCE      1000
1  2017Q1              MEDICAL      1000
2  2017Q3               FAMILY      5000
3  2018Q1       ART_AND_DESIGN    500000
4  2018Q2       ART_AND_DESIGN  50000000
5  2018Q3       ART_AND_DESIGN   5000000
6  2018Q3               FAMILY       100

或者，如果需要按季度和类别汇总安装并获取最大安装的查询者，则使用:

q = (pd.to_datetime(df['Month'] + df['Year'].astype(str), format='%B%Y')
       .dt.to_period('Q').rename('Quarter'))

df1 = df.groupby([q,'Category'])['Installs'].sum().reset_index()
print (df1)
  Quarter             Category  Installs
0  2015Q1  BOOKS_AND_REFERENCE      1000
1  2017Q1              MEDICAL      1000
2  2017Q3               FAMILY      5000
3  2018Q1       ART_AND_DESIGN    510000
4  2018Q2       ART_AND_DESIGN  50100000
5  2018Q3       ART_AND_DESIGN   5000000
6  2018Q3               FAMILY       100
7  2018Q3            LIFESTYLE  10000000

df2 = df1.loc[df1.groupby('Category')['Installs'].idxmax()]
print (df2)
  Quarter             Category  Installs
4  2018Q2       ART_AND_DESIGN  50100000
0  2015Q1  BOOKS_AND_REFERENCE      1000
2  2017Q3               FAMILY      5000
7  2018Q3            LIFESTYLE  10000000
1  2017Q1              MEDICAL      1000