Pandas top n values in each group

Question

I have a dataframe like

item      date       hour     value
  a         4         12       123
  a         6         11        54
  b         1          7       146
  c         8          1        97
  a         9          5        10
  c         4          5       114
  b         1          7       200
...       ...        ...       ...

and I want to keep the top 10 item by value (discard the rest is ok), regardless any other column. They are not sorted.

Following my input example, and as I didn't write enough to get 10 from every item, the expected output would be something like this if I want the top 1:

item      date       hour     value
  a         4         12       123
  c         4          5       114
  b         1          7       200
...       ...        ...       ...

I've seen this answer but I'm not sure how to tell pandas to take value for the calculation.

Can you provide an example of an input data frame and the expected output? — SEDaradji, Oct 07 '19 at 15:41
`df.groupby('item')['value'].nlargest(10)` the many dupes cover some other options — user3483203, Oct 07 '19 at 15:45

score 11 · Accepted Answer · answered Oct 07 '19 at 15:43

11

You can sort_values by both ['item', 'value'] and then groupby.head:

df.sort_values(['item', 'value'], ascending=False).groupby('item').head(10)

Or with nlargest:

df.groupby('item').value.nlargest(10).reset_index()

answered Oct 07 '19 at 15:43

yatu

86,083
12
84
139

Pandas top n values in each group

1 Answers1