python - Pandas - 多轴上的 groupby() 上的 cumsum()

Question

我是 Pandas 新手。 我有一个数据集是赛马结果。示例如下:

RaceID  RaceDate    RaceMeet   Position Horse       Jockey          Trainer        RaceLength race win  HorseWinPercentage 
446252  01/01/2008  Southwell (AW)  1   clear reef  tom mclaughlin  jane chapple-hyam   3101    1   1   0
447019  14/01/2008  Southwell (AW)  5   clear reef  tom mclaughlin  jane chapple-hyam   2654    1   0   100
449057  21/02/2008  Southwell (AW)  2   clear reef  tom mclaughlin  jane chapple-hyam   3101    1   0   50
463805  26/08/2008  Chelmsford (AW) 6   clear reef  tom mclaughlin  jane chapple-hyam   3080    1   0   33.33333333
469220  27/11/2008  Chelmsford (AW) 3   clear reef  tom mclaughlin  jane chapple-hyam   3080    1   0   25
470195  11/12/2008  Chelmsford (AW) 5   clear reef  tom mclaughlin  jane chapple-hyam   3080    1   0   20
471052  26/12/2008  Wolhampton (AW) 1   clear reef  andrea atzeni   jane chapple-hyam   2690    1   1   16.66666667
471769  07/01/2009  Wolhampton (AW) 6   clear reef  ian mongan      jane chapple-hyam   2690    1   0   28.57142857
472137  13/01/2009  Chelmsford (AW) 2   clear reef  jamie spencer   jane chapple-hyam   3080    1   0   25
472213  20/01/2009  Southwell (AW)  5   clear reef  jamie spencer   jane chapple-hyam   2654    1   0   22.22222222
476595  25/03/2009  Kempton (AW)    4   clear reef  pat cosgrave    jane chapple-hyam   2639    1   0   20
477674  08/04/2009  Kempton (AW)    5   clear reef  pat cosgrave    jane chapple-hyam   2639    1   0   18.18181818
479098  21/04/2009  Kempton (AW)    3   clear reef  andrea atzeni   jane chapple-hyam   2639    1   0   16.66666667
492913  14/11/2009  Wolhampton (AW) 1   clear reef  andrea atzeni   jane chapple-hyam   3639    1   1   15.38461538
493720  25/11/2009  Kempton (AW)    3   clear reef  andrea atzeni   jane chapple-hyam   3518    1   0   21.42857143
495863  29/12/2009  Southwell (AW)  1   clear reef  shane kelly     jane chapple-hyam   3101    1   1   20

我希望能够通过 groupby() 多个轴来计算获胜次数并创建特定赛道和长度的组合获胜百分比或结果。

当我只需要按单个轴进行分组时 - 它效果很好:

df['horse_win_count'] = df.groupby(['Horse'])['win'].cumsum()
df['horse_race_count'] = df.groupby(['Horse'])['race'].cumsum()
df['HorseWinPercentage2'] = df['horse_win_count'] / df['horse_race_count'] * 100
df['HorseWinPercentage'] = df.groupby('Horse')['HorseWinPercentage2'].shift(+1)

但是，当我需要对多个轴进行分组时，我会得到一些非常奇怪的结果。

例如，我要为特定骑师骑特定训练师的马时创建获胜百分比 - groupby([‘Jockey’,’Trainer’])。然后我需要知道每行(比赛)变化的百分比。

df['jt_win_count'] = df.groupby(['Jockey','Trainer'])['win'].cumsum()
df['jt_race_count'] = df.groupby(['Jockey','Trainer'])['race'].cumsum()
df['JTWinPercentage2'] = df['jt_win_count'] / df['jt_race_count'] * 100
df['JTWinPercentage'] = df.groupby(['Jockey','Trainer'])['JTWinPercentage2'].shift(+1)
df['JTWinPercentage'].fillna(0, inplace=True)

或者我想计算一匹马在该路线和距离上获胜的次数。所以我需要 groupby(['Horse', 'RaceMeet','RaceLength']):

df['CD'] = df.groupby([‘RaceMeet’,’RaceLength’,’Horse’])[‘win’].cumsum()
df['CD'] = df.groupby(["RaceMeet","RaceLength","Horse"]).shift(+1)

我得到了 10 到 1000 的结果。

如何在按多个条目分组时按多个轴进行分组、进行计算并将结果向后移一个条目？

您能解释一下为什么我上面的代码不起作用吗？就像我说的，我是 Pandas 新手，并且热衷于学习。

干杯。

Answer 1

问题已被提出:Pandas DataFrame Groupby two columns and get counts和这里 python pandas groupby() result

我真的不知道你的目标是什么。 我想您应该首先添加另一列，其中包含要分组的新参数。例如: df['jockeyTrainer']=df.loc['Jockey']+df.loc['Trainer'] 然后你可以用它来分组。或者您可以按照链接中的信息进行操作。