I have a pandas dataframe and I wish to divide it to 3 separate sets. I know that using train_test_split from sklearn.cross_validation
, one can divide the data in two sets (train and test). However, I couldn't find any solution about splitting the data into three sets. Preferably, I'd like to have the indices of the original data.
I know that a workaround would be to use train_test_split
two times and somehow adjust the indices. But is there a more standard / built-in way to split the data into 3 sets instead of 2?
Numpy solution. We will shuffle the whole dataset first (df.sample(frac=1, random_state=42)
) and then split our data set into the following parts:
In [305]: train, validate, test = \
np.split(df.sample(frac=1, random_state=42),
[int(.6*len(df)), int(.8*len(df))])
In [306]: train
Out[306]:
A B C D E
0 0.046919 0.792216 0.206294 0.440346 0.038960
2 0.301010 0.625697 0.604724 0.936968 0.870064
1 0.642237 0.690403 0.813658 0.525379 0.396053
9 0.488484 0.389640 0.599637 0.122919 0.106505
8 0.842717 0.793315 0.554084 0.100361 0.367465
7 0.185214 0.603661 0.217677 0.281780 0.938540
In [307]: validate
Out[307]:
A B C D E
5 0.806176 0.008896 0.362878 0.058903 0.026328
6 0.145777 0.485765 0.589272 0.806329 0.703479
In [308]: test
Out[308]:
A B C D E
4 0.521640 0.332210 0.370177 0.859169 0.401087
3 0.333348 0.964011 0.083498 0.670386 0.169619
[int(.6*len(df)), int(.8*len(df))]
- is an indices_or_sections
array for numpy.split().
Here is a small demo for np.split()
usage - let's split 20-elements array into the following parts: 80%, 10%, 10%:
In [45]: a = np.arange(1, 21)
In [46]: a
Out[46]: array([ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20])
In [47]: np.split(a, [int(.8 * len(a)), int(.9 * len(a))])
Out[47]:
[array([ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]),
array([17, 18]),
array([19, 20])]