How to select the last column of dataframe

Question

I have done some searching for the answer to this question, but all I can figure out is this:

df[df.columns[len(df.columns)-1]]

which to me seems unweildy, and un-pythonic (and slow?).

What is the easiest way to select the data for the last column in a pandas dataframe without specifying the name of the column?

score 205 · Accepted Answer · edited Sep 18 '20 at 14:24

205

Use iloc and select all rows (:) against the last column (-1):

df.iloc[:,-1:]

edited Sep 18 '20 at 14:24

svet

11,078
1
15
26

answered Oct 20 '16 at 04:36

Zeugma

31,231
9
69
81

6

I tried to do that and got an error saying "too many indexers" – user287474 Apr 28 '18 at 02:12
1

@user287474 that'll happen if you're operating over a `Series` - make sure you're using a `DataFrame` – Max Ghenis Mar 01 '21 at 00:27

score 49 · Answer 2 · answered Oct 20 '16 at 03:17

49

Somewhat similar to your original attempt, but more Pythonic, is to use Python's standard negative-indexing convention to count backwards from the end:

df[df.columns[-1]]

answered Oct 20 '16 at 03:17

jez

14,867
5
37
64

will this work? ```df[df.columns[-1] >= 10]``` it gives me "not supported between instances of 'str' and 'int'" – Mike_Leigh May 07 '21 at 14:09
@Mike_Leigh your comparison shouldn’t be happening inside the bracket – jez May 08 '21 at 12:14

score 30 · Answer 3 · edited Mar 24 '20 at 17:47

These are few things which will help you in understanding everything... using iloc

In iloc, [initial row:ending row, initial column:ending column]

case 1: if you want only last column --- df.iloc[:,-1] & df.iloc[:,-1:] this means that you want only the last column...

case 2: if you want all columns and all rows except the last column --- df.iloc[:,:-1] this means that you want all columns and all rows except the last column...

case 3: if you want only last row --- df.iloc[-1:,:] & df.iloc[-1,:] this means that you want only the last row...

case 4: if you want all columns and all rows except the last row --- df.iloc[:-1,:] this means that you want all columns and all rows except the last column...

case 5: if you want all columns and all rows except the last row and last column --- df.iloc[:-1,:-1] this means that you want all columns and all rows except the last column and last row...

score 9 · Answer 4 · edited Sep 11 '20 at 13:39

Just to add to @Anshul Singh Suryan's answer:

When we split the dataframe to just get the last column:

If we split like:

y = df.iloc[:,-1:] - y remains a dataframe

However, if we split like

y = df.iloc[:,-1] - y becomes a Series.

This is a notable difference that I've found in the two approaches. If you don't care about the resultant type, you can use either of the two. Otherwise you need to take care of the above findings.

This is applicable for any number of rows you want to extract and not just the last row. For example, if you want last n number of rows of a dataframe, where n is any integer less than or equal to the number of columns present in the dataframe, then you can easily do the following:

y = df.iloc[:,n:]

Replace n by the number of columns you want. Same is true for rows as well.

score 7 · Answer 5 · answered Jun 24 '17 at 08:36

7

The question is: how to select the last column of a dataframe ? Appart @piRSquared, none answer the question.

the simplest way to get a dataframe with the last column is:

df.iloc[ :, -1:]

answered Jun 24 '17 at 08:36

alEx

193
6
12

score 3 · Answer 6 · answered Oct 20 '16 at 06:13

3

df.T.iloc[-1]

df.T.tail(1)

pd.Series(df.values[:, -1], name=df.columns[-1])

answered Oct 20 '16 at 06:13

piRSquared

285,575
57
475
624

1

I don't see any value of building the Series explicitly out of the numpy values extraction compared to a direct call to iloc – Zeugma Oct 20 '16 at 12:42
@Boud, There are many ways to do this. I was simply trying to dump a bunch down. Admittedly, the third option is the weakest. – piRSquared Oct 20 '16 at 12:58

score 1 · Answer 7 · answered Mar 24 '17 at 18:28

1

This is another way to do it. I think maybe a little more general:

df.ix[:,-1]

answered Mar 24 '17 at 18:28

user28929304981

29
2

1

In this case, your answer is not recommended because the question asks explicitly for the *last* column which is index location based. `ix` mixes label and index location accessor. See [here](http://stackoverflow.com/questions/31593201/pandas-iloc-vs-ix-vs-loc-explanation) for more. – pansen Mar 24 '17 at 19:10
4

Indeed, in addition, `ix` is deprecated by now. – Archie Apr 12 '18 at 07:59

score 0 · Answer 8 · answered Jun 25 '22 at 19:18

0

Alternatively you can also use take:

df.take([-1], axis=1)

answered Jun 25 '22 at 19:18

rachwa

1,805
1
14
17

How to select the last column of dataframe

8 Answers8

Linked

Related