44

I'm wondering if there is a simpler, memory efficient way to select a subset of rows and columns from a pandas DataFrame.

For instance, given this dataframe:

df = DataFrame(np.random.rand(4,5), columns = list('abcde'))
print df

          a         b         c         d         e
0  0.945686  0.000710  0.909158  0.892892  0.326670
1  0.919359  0.667057  0.462478  0.008204  0.473096
2  0.976163  0.621712  0.208423  0.980471  0.048334
3  0.459039  0.788318  0.309892  0.100539  0.753992

I want only those rows in which the value for column 'c' is greater than 0.5, but I only need columns 'b' and 'e' for those rows.

This is the method that I've come up with - perhaps there is a better "pandas" way?

locs = [df.columns.get_loc(_) for _ in ['a', 'd']]
print df[df.c > 0.5][locs]

          a         d
0  0.945686  0.892892

My final goal is to convert the result to a numpy array to pass into an sklearn regression algorithm, so I will use the code above like this:

training_set = array(df[df.c > 0.5][locs])

... and that peeves me since I end up with a huge array copy in memory. Perhaps there's a better way for that too?

Improve this question

3 Answers 3

Reset to default
71

Use its value directly:

In [79]: df[df.c > 0.5][['b', 'e']].values
Out[79]: 
array([[ 0.98836259,  0.82403141],
       [ 0.337358  ,  0.02054435],
       [ 0.29271728,  0.37813099],
       [ 0.70033513,  0.69919695]])
Improve this answer
Sign up to request clarification or add additional context in comments.

4 Comments

I didn't know about the .values attribute. Very nice! Also, slightly cleaner since you eliminated the single quotes & brackets and instead used df.c directly.
nice but how is it different from as_matrix though?
Just an update, as I was just wondering about the difference between as_matrix and .values (since I only use .values). It turns out that as_matrix is only provided for backwards compatibility, and that it is recommended to use .values instead. See pandas.pydata.org/pandas-docs/version/0.17.1/generated/…
Sadly .values doesn't allow filtering or ordering the columns.
16

Perhaps something like this for the first problem, you can simply access the columns by their names:

>>> df = pd.DataFrame(np.random.rand(4,5), columns = list('abcde'))
>>> df[df['c']>.5][['b','e']]
          b         e
1  0.071146  0.132145
2  0.495152  0.420219

For the second problem:

>>> df[df['c']>.5][['b','e']].values
array([[ 0.07114556,  0.13214495],
       [ 0.49515157,  0.42021946]])
Improve this answer

Comments

9

.loc accept row and column selectors simultaneously (as do .ix/.iloc FYI) This is done in a single pass as well.

In [1]: df = DataFrame(np.random.rand(4,5), columns = list('abcde'))

In [2]: df
Out[2]: 
          a         b         c         d         e
0  0.669701  0.780497  0.955690  0.451573  0.232194
1  0.952762  0.585579  0.890801  0.643251  0.556220
2  0.900713  0.790938  0.952628  0.505775  0.582365
3  0.994205  0.330560  0.286694  0.125061  0.575153

In [5]: df.loc[df['c']>0.5,['a','d']]
Out[5]: 
          a         d
0  0.669701  0.451573
1  0.952762  0.643251
2  0.900713  0.505775

And if you want the values (though this should pass directly to sklearn as is); frames support the array interface

In [6]: df.loc[df['c']>0.5,['a','d']].values
Out[6]: 
array([[ 0.66970138,  0.45157274],
       [ 0.95276167,  0.64325143],
       [ 0.90071271,  0.50577509]])
Improve this answer

2 Comments

Most elegant. What's the difference between .ix and .loc?
loc will not attempt to use a number (eg 1) as a positional argument at all (and will raise instead); see main pandas docs / selecting data

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.