Taking maximum value of only one column by “groupby” in pandas

Question

I have a dataframe with 10 columns:

id        date         value
1233     2014-10-3     1.123123
3412     2015-05-31    2.123123
3123     2015-05-31    5.6234234
3123     2013-03-21    5.6234222
3412     2014-11-21    4.776666
5121     2015-08-22    5.234234

I want to group by id column and take the latest date . But I don't want to take the maximum of value column. I want to take the value fo such row, that belongs to the maximum date.

pd.groupby('id').max() doesn't work. How can I solve it?

The most important thing, that I want to keep all columns in my dataset.

Answer 1

You can use boolean indexing to select the max date in a group and return that row per group:

df.groupby('id').apply(lambda x: x.loc[x.date == x.date.max(),['date','value']])

Or, use idxmax to select the index of that maximum value in each group:

df.groupby('id').apply(lambda x: x.loc[x.date.idxmax(),['date','value']]).reset_index()

Output:

     id       date     value
0  1233 2014-10-03  1.123123
1  3123 2015-05-31  5.623423
2  3412 2015-05-31  2.123123
3  5121 2015-08-22  5.234234

Answer 2

Or you can simply using sort_value then first

df.sort_values(['date', 'value'], ascending=[False, True]).groupby('id').first()

Out[480]: 
           date     value
id                       
1233 2014-10-03  1.123123
3123 2015-05-31  5.623423
3412 2015-05-31  2.123123
5121 2015-08-22  5.234234

Answer 3

You could sort by date, then keep only the first appearance of each id.

df = df.sort_values('date', ascending=False)

most_recent = df.drop_duplicates('id', keep='first')

most_recent

Out: 
     id        date     value
0  5121  2015-08-22  5.234234
1  3412  2015-05-31  2.123123
2  3123  2015-05-31  5.623423
4  1233   2014-10-3  1.123123

Answer 4

如果要返回包含最大日期的整行，则需要使用idxmax ：

result_row = df.iloc[df['date'].idxmax()]