从二维数组中选择行而不是在另一个二维数组中

Question

i want to choose randomly 25% rows of a 2D array the return the rows i choosed and not choosed also.

Example:

cols = [['republican' 'n' 'y' 'n' 'n' 'y']
 ['republican' 'n' 'y' 'y' 'n' 'n']
 ['democrat' 'y' 'y' 'y' 'n' 'n']
['republican' 'n' 'y' 'y' 'n' 'y']]

new = [['democrat' 'y' 'y' 'y' 'n' 'n']]

rem = [['republican' 'n' 'y' 'n' 'n' 'y']
 ['republican' 'n' 'y' 'y' 'n' 'n']
['republican' 'n' 'y' 'y' 'n' 'y']]

here is my code:

def splitRandom(cols, percent):
    rng = np.random.default_rng()
    new = cols[np.random.choice(cols.shape[0], int(cols.shape[0]*(percent/100)), replace=False)]
    rem = #help here
    return new,rem

i tried to use setdiff1d but didn't work and i don't know how to mask it.

Answer 1

I modified your code slightly:

def splitRandom(cols, percent):
    rng = np.random.default_rng()
    choise = np.random.choice(cols.shape[0], int(cols.shape[0] * (percent / 100)),replace=False)
    new = cols[choise]
    rem = np.delete(cols, choise, axis=0)
    return new,rem

edit: I assume cols is a numpy array

Answer 2

Here is an alternate approach that avoids modification of the input array:

def splitRandom(cols, percent):
    rng = np.random.default_rng()
    choise = np.random.choice(cols.shape[0], int(cols.shape[0] * (percent / 100)),replace=False)
    mask = np.zeros(cols.shape[0], dtype=bool)
    mask[choise] = True
    return cols[mask], cols[~mask]

Note that this sort of operation is common enough that machine learning frameworks often have built-in methods to do it efficiently; for example, here is the equivalent using sklearn.model_selection.train_test_split :

from sklearn.model_selection import train_test_split
train_test_split(cols, train_size=0.25)