How to rank rows by id in Pandas Python

问题

I have a Dataframe like this:

id     points1    points2
1        44          53
1        76          34
1        63          66
2        23          34 
2        44          56

I want output like this:

id     points1    points2     points1_rank     points2_rank
1        44          53            3                2
1        76          34            1                3
1        63          66            2                1
2        23          79            2                1
2        44          56            1                2

Basically, I want to groupby('id'), and find the rank of each column with same id.

I tried this:

features = ["points1","points2"]
df = pd.merge(df, df.groupby('id')[features].rank().reset_index(), suffixes=["", "_rank"], how='left', on=['id'])

But I get keyerror 'id'

回答1:

You need to use ascending=False inside rank

df.join(df.groupby('id')['points1', 'points2'].rank(ascending=False).astype(int).add_suffix('_rank'))


+---+----+---------+---------+--------------+--------------+
|   | id | points1 | points2 | points1_rank | points2_rank |
+---+----+---------+---------+--------------+--------------+
| 0 |  1 |      44 |      53 |            3 |            2 |
| 1 |  1 |      76 |      34 |            1 |            3 |
| 2 |  1 |      63 |      66 |            2 |            1 |
| 3 |  2 |      23 |      34 |            2 |            2 |
| 4 |  2 |      44 |      56 |            1 |            1 |
+---+----+---------+---------+--------------+--------------+

回答2:

Use join with remove reset_index and for change columns names add add_suffix:

features = ["points1","points2"]
df = df.join(df.groupby('id')[features].rank(ascending=False).add_suffix('_rank').astype(int))
print (df)
   id  points1  points2  points1_rank  points2_rank
0   1       44       53             3             2
1   1       76       34             1             3
2   1       63       66             2             1
3   2       23       34             2             2
4   2       44       56             1             1

来源：https://stackoverflow.com/questions/54017968/how-to-rank-rows-by-id-in-pandas-python