[PYTHON] Use iter trat for multi-label stratified CV

Since machine learning models tend to overfit with training data, we often divide the data at hand into training data and test data and perform performance evaluation (validation). The explanation about validation of scikit-learn is easy to understand about various methods of this division. https://scikit-learn.org/stable/modules/cross_validation.html#cross-validation

1. In case of Stratified K Fold of scikit-learn

When doing classification, I want to divide so that the distribution of correct labels is the same, so I think that Stratified K Fold of scikit-learn is often used.

python

import numpy as np
from sklearn.model_selection import StratifiedKFold

X = np.array([[1, 2], [3, 4], [1, 2], [3, 4]])
y = np.array([0, 0, 1, 1])
skf = StratifiedKFold(n_splits=2)
for train_index, test_index in skf.split(X, y):
    print("TRAIN:", train_index, "TEST:", test_index)
    X_train, X_test = X[train_index], X[test_index]
    y_train, y_test = y[train_index], y[test_index]

result

TRAIN: [1 3] TEST: [0 2]
TRAIN: [0 2] TEST: [1 3]

This is sufficient if you have only one label, but it does not support multiple labels.

python

import numpy as np
from sklearn.model_selection import StratifiedKFold

X = np.array([[1,2], [3,4], [1,2], [3,4], [1,2], [3,4], [1,2], [3,4]])
y = np.array([[0,0], [0,0], [0,1], [0,1], [1,1], [1,1], [1,0], [1,0]])
skf = StratifiedKFold(n_splits=2)

for train_index, test_index in skf.split(X, y):
    print("TRAIN:", train_index, "TEST:", test_index)
    X_train, X_test = X[train_index], X[test_index]
    y_train, y_test = y[train_index], y[test_index]

result

ValueError: Supported target types are: ('binary', 'multiclass'). Got 'multilabel-indicator' instead.

2. For iterstrat

iterstrat supports multiple labels https://github.com/trent-b/iterative-stratification

Installation

terminal

pip install iterative-stratification

How to use (when not shuffle)

python

from iterstrat.ml_stratifiers import MultilabelStratifiedKFold
import numpy as np

X = np.array([[1,2], [3,4], [1,2], [3,4], [1,2], [3,4], [1,2], [3,4]])
y = np.array([[0,0], [0,0], [0,1], [0,1], [1,1], [1,1], [1,0], [1,0]])

mskf = MultilabelStratifiedKFold(n_splits=2, shuffle=True, random_state=0)

for train_index, test_index in mskf.split(X, y):
    print("TRAIN:", train_index, "TEST:", test_index)
    X_train, X_test = X[train_index], X[test_index]
    y_train, y_test = y[train_index], y[test_index]

result

TRAIN: [0 3 4 6] TEST: [1 2 5 7]
TRAIN: [1 2 5 7] TEST: [0 3 4 6]

How to use (when shuffle)

python

from iterstrat.ml_stratifiers import MultilabelStratifiedShuffleSplit
import numpy as np

X = np.array([[1,2], [3,4], [1,2], [3,4], [1,2], [3,4], [1,2], [3,4]])
y = np.array([[0,0], [0,0], [0,1], [0,1], [1,1], [1,1], [1,0], [1,0]])

msss = MultilabelStratifiedShuffleSplit(n_splits=3, test_size=0.5, random_state=0)

for train_index, test_index in msss.split(X, y):
   print("TRAIN:", train_index, "TEST:", test_index)
   X_train, X_test = X[train_index], X[test_index]
   y_train, y_test = y[train_index], y[test_index]

result

TRAIN: [1 2 5 7] TEST: [0 3 4 6]
TRAIN: [2 3 6 7] TEST: [0 1 4 5]
TRAIN: [1 2 5 6] TEST: [0 3 4 7]

How to use (when bootstrap)

python

from iterstrat.ml_stratifiers import RepeatedMultilabelStratifiedKFold
import numpy as np

X = np.array([[1,2], [3,4], [1,2], [3,4], [1,2], [3,4], [1,2], [3,4]])
y = np.array([[0,0], [0,0], [0,1], [0,1], [1,1], [1,1], [1,0], [1,0]])

rmskf = RepeatedMultilabelStratifiedKFold(n_splits=2, n_repeats=2, random_state=0)

for train_index, test_index in rmskf.split(X, y):
   print("TRAIN:", train_index, "TEST:", test_index)
   X_train, X_test = X[train_index], X[test_index]
   y_train, y_test = y[train_index], y[test_index]

result

TRAIN: [0 3 4 6] TEST: [1 2 5 7]
TRAIN: [1 2 5 7] TEST: [0 3 4 6]
TRAIN: [0 1 4 5] TEST: [2 3 6 7]
TRAIN: [2 3 6 7] TEST: [0 1 4 5]