[英]Pandas find the maximum in each group that satisfy a condition
Sorry if this has been asked before, could not find an exact question. 对不起,如果之前有人问过,找不到确切的问题。
I am looking for the most efficient way in Pandas to do the following operation: 我正在寻找Pandas中最有效的方法来执行以下操作:
Lets say we have the following table: 可以说我们有下表:
ID SUB_ID COND
1 101 1 1
2 101 2 1
3 101 3 1
4 102 1 1
5 102 2 0
6 103 1 0
7 103 2 0
8 103 3 0
9 103 4 0
Basically, for each "ID" we want to get the largest "SUB_ID", given that the COND is 1 . 基本上,对于每个“ID”,我们希望得到最大的“SUB_ID”, 假设COND为1 。 Ideally we would want to add this value to each row of that ID as a new column.
理想情况下,我们希望将此值作为新列添加到该ID的每一行。 If no row of that ID fulfills the condition, we would like to add a 0 (instead of null)
如果该ID的行没有满足条件,我们想添加0(而不是null)
Resulting dataframe would be: 结果数据框将是:
ID SUB_ID COND MAX_SUB_ID
1 101 1 1 3
2 101 2 1 3
3 101 3 1 3
4 102 1 1 1
5 102 2 0 1
6 103 1 0 0
7 103 2 0 0
8 103 3 0 0
9 103 4 0 0
Best way I can come up with right now is selecting only the rows where COND=1, then doing a groupby on this dataframe to get the max sub id, and then joining it back to the main dataframe. 我现在能想出的最佳方法是仅选择COND = 1的行,然后在此数据帧上执行groupby以获取最大子ID,然后将其连接回主数据帧。 After this I can change the null back to 0.
在此之后,我可以将null更改回0。
df_true = df[df['COND']==1]
max_subid_true=df_true['SUB_ID'].groupby(df_true['ID']).max()
joined_df = df.merge(pd.DataFrame(max_subid_true),how='left',left_on='ID',right_index=True)
joined_df.loc[pd.isnull(joined_df['SUB_ID_y']),'SUB_ID_y']=0
Any ideas on doing this differently? 有什么不同的想法吗?
df.assign(MAX_SUB_ID=df.SUB_ID.mul(df.COND).groupby(df.ID).transform('max'))
ID SUB_ID COND MAX_SUB_ID
1 101 1 1 3
2 101 2 1 3
3 101 3 1 3
4 102 1 1 1
5 102 2 0 1
6 103 1 0 0
7 103 2 0 0
8 103 3 0 0
9 103 4 0 0
caveats 注意事项
SUB_ID
is always positive SUB_ID
始终为正 COND
is always 1
or 0
COND
始终为1
或0
alternative (with less caveats) 替代方案(减少警告)
but less fun 但不那么有趣
df.assign(MAX_SUB_ID=df.ID.map(df.query('COND == 1').groupby('ID').SUB_ID.max()) \
.fillna(0).astype(int))
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.