extracting equal number of samples for each class in a unbalanced samples scenario

Question

I have 2 classes, A and B. Each class has an unbalanced number of samples to each other, say 500 to class A and 1000 to class B.

Is there a way to extract balanced number of samples for each class, say 300 for class A and B using scikit learn or any of the Numpy functions.

The samples are the first 5 columns and the labels/classes is the final column

1 2 3 4 5 1
2 3 4 2 3 1
4 0 5 4 3 1
4 5 9 2 4 2
5 9 5 3 9 2

What I would like to be done is have an equal amount of both classes in my final pick:

2 3 4 2 3 1
4 0 5 4 3 1
4 5 9 2 4 2
5 9 5 3 9 2

Answer 1

You can use sample() method with the same value for n parameter, if you can use Pandas

Demo:

In [364]: df1
Out[364]:
          a         b         c         d
0  0.774496  0.852985  0.257568  0.223773
1  0.630460  0.203675  0.305280  0.965628
2  0.408746  0.939827  0.801505  0.343216
3  0.578582  0.541716  0.451810  0.173890
4  0.210301  0.600485  0.184326  0.035092
5  0.583564  0.164262  0.958537  0.943357

In [365]: df2
Out[365]:
          a         b         c         d
0  0.340624  0.755825  0.569149  0.543630
1  0.004056  0.463891  0.556861  0.778607
2  0.171046  0.293104  0.317514  0.831424
3  0.370028  0.566356  0.895919  0.440559
4  0.148569  0.485086  0.299789  0.274720
5  0.137273  0.085598  0.874845  0.917356
6  0.356898  0.781540  0.091851  0.173430
7  0.495949  0.613337  0.512104  0.137251

In [366]: df1.sample(n=5)
Out[366]:
          a         b         c         d
3  0.578582  0.541716  0.451810  0.173890
4  0.210301  0.600485  0.184326  0.035092
1  0.630460  0.203675  0.305280  0.965628
0  0.774496  0.852985  0.257568  0.223773
5  0.583564  0.164262  0.958537  0.943357

In [367]: df2.sample(n=5)
Out[367]:
          a         b         c         d
2  0.171046  0.293104  0.317514  0.831424
5  0.137273  0.085598  0.874845  0.917356
6  0.356898  0.781540  0.091851  0.173430
3  0.370028  0.566356  0.895919  0.440559
0  0.340624  0.755825  0.569149  0.543630