数据透视表

使用 pandas 重塑数据

Maria Eugenia Inzaugarat

Data Scientist

pivot 方法的局限

another_fifa.head()

                 name    variable  metric_system  imperial_system
0   Cristiano Ronaldo      weight             83           183.00
1            J. Oblak      weight             87           191.00
2   Cristiano Ronaldo      height            187             6.13
3     J. Oblak             height            188             6.16
4   Cristiano Ronaldo      height            187             6.14

another_fifa.pivot(index="name", columns="variable")

Traceback (most recent call last):
   ValueError: Index contains duplicate entries, cannot reshape

pivot 方法的局限

通用旋转
索引/列对必须唯一
不能聚合值

数据透视表

一个 DataFrame，用统计量汇总更大 DataFrame 的数据

显示汇总统计的 DataFrame

数据透视表

长格式的 DataFrame

数据透视表

箭头从长表到宽表并含汇总统计

pivot_table 的调用

数据透视表

长表与汇总 DataFrame，含高亮列与索引

带 index 与 columns 参数的 pivot 调用

数据透视表

长表与汇总 DataFrame，含高亮列与值

带 values 与聚合函数参数的 pivot 调用

数据透视表

another_fifa.pivot_table(index="name", columns="variable", aggfunc="mean")

                     metric_system     imperial_system       
         variable   height  weight     height   weight
             name                                                         
Cristiano Ronaldo      187      83      6.135    183.0
         J. Oblak      188      87      6.160    191.0

分层索引

fifa_players.head(6)

      first     last   movement   overall  attacking
0    Lionel    Messi   shooting        92         70
1 Cristiano  Ronaldo   shooting        93         89
2    Lionel    Messi    passing        92         92
3 Cristiano  Ronaldo    passing        82         83
4    Lionel    Messi    passing        96         88
5 Cristiano  Ronaldo    passing        89         84

分层索引

fifa_players.head(6)

      first     last   movement   overall  attacking
0    Lionel    Messi   shooting        92         70
1 Cristiano  Ronaldo   shooting        93         89
2    Lionel    Messi    passing        92         92
3 Cristiano  Ronaldo    passing        82         83
4    Lionel    Messi    passing        96         88
5 Cristiano  Ronaldo    passing        89         84

fifa_players.pivot_table(index=                 , columns="movement", values=                        , aggfunc=     )

分层索引

fifa_players.head(6)

      first     last   movement   overall  attacking
0    Lionel    Messi   shooting        92         70
1 Cristiano  Ronaldo   shooting        93         89
2    Lionel    Messi    passing        92         92
3 Cristiano  Ronaldo    passing        82         83
4    Lionel    Messi    passing        96         88
5 Cristiano  Ronaldo    passing        89         84

fifa_players.pivot_table(index=["first", "last"], columns="movement", values=                        , aggfunc=     )

分层索引

fifa_players.head(6)

      first     last   movement   overall  attacking
0    Lionel    Messi   shooting        92         70
1 Cristiano  Ronaldo   shooting        93         89
2    Lionel    Messi    passing        92         92
3 Cristiano  Ronaldo    passing        82         83
4    Lionel    Messi    passing        96         88
5 Cristiano  Ronaldo    passing        89         84

fifa_players.pivot_table(index=["first", "last"], columns="movement", values=["overall", "attacking"], aggfunc="max")

                           attacking           overall
          movement  passing shooting  passing shooting
    first     last                
Cristiano  Ronaldo       84       89       89       93
   Lionel    Messi       92       70       96       92

合计行列（margins）

fifa_players.pivot_table(index=["first", "last"], columns="movement", aggfunc="count",             )

合计行列（margins）

fifa_players.pivot_table(index=["first", "last"], columns="movement", aggfunc="count", margins=True)

                                attacking                  overall
          movement  passing shooting  All    passing shooting  All
    First     Last                
Cristiano  Ronaldo        2        1    3          2        1    3
   Lionel    Messi        2        1    3          2        1    3
      All                 4        2    6          4        2    6

用 pivot 还是 pivot_table？

"每个索引/列对是否有多个值？"

"结果需要多级索引吗？"

"需要对大型 DataFrame 做汇总统计吗？"

是的！请用 .pivot_table()

让我们来练习！

使用 pandas 重塑数据