Generate an imbalanced dataset

An illustration of using the generate_imbalance_data() function to create an imbalanced dataset.

# Authors: Zhining Liu <zhining.liu@outlook.com>
# License: MIT

print(__doc__)

from imbens.datasets import generate_imbalance_data
from imbens.utils._plot import plot_2Dprojection_and_cardinality
from collections import Counter

Generate the dataset

X_train, X_test, y_train, y_test = generate_imbalance_data(
    n_samples=1000,
    weights=[0.7, 0.2, 0.1],
    test_size=0.5,
    kwargs={'n_informative': 3},
)

print("Train class distribution: ", Counter(y_train))
print("Test class distribution:  ", Counter(y_test))

Train class distribution:  Counter({np.int64(0): 344, np.int64(1): 103, np.int64(2): 53})
Test class distribution:   Counter({np.int64(0): 343, np.int64(1): 103, np.int64(2): 54})

Plot the generated (training) data

plot_2Dprojection_and_cardinality(X_train, y_train)

(<Figure size 1000x400 with 2 Axes>, (<Axes: title={'center': 'Dataset (2D projection by KernelPCA)'}>, <Axes: title={'center': 'Class Distribution'}, xlabel='Class'>))