predict_proba() with LinearSVC as classifier about scikit-multilearn HOT 1 OPEN

from sklearn.svm import LinearSVC
from sklearn.calibration import CalibratedClassifierCV
from skmultilearn.problem_transform import BinaryRelevance
from sklearn.datasets import make_multilabel_classification
from sklearn.model_selection import train_test_split


X, y = make_multilabel_classification(n_samples=1000, n_features=20, n_classes=5, n_labels=2, random_state=42)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)

# Initialize the LinearSVC classifier
base_clf = LinearSVC()

# Calibrate the classifier
calibrated_clf = CalibratedClassifierCV(base_clf)

# Use the calibrated classifier with for instance BinaryRelevance from scikit-multilearn
classifier = BinaryRelevance(classifier=calibrated_clf)

# Train the classifier
classifier.fit(X_train, y_train)

# Now you can use predict_proba
probabilities = classifier.predict_proba(X_test)

print(probabilities)

import numpy as np

# Convert the sparse matrix to a dense format for manipulation
prob_dense = probabilities.toarray()

# Normalize the probabilities row-wise
prob_normalized = prob_dense / np.sum(prob_dense, axis=1, keepdims=True)

# Check the sum again to ensure they are normalized
prob_sum_normalized = prob_normalized.sum(axis=1)

prob_normalized, prob_sum_normalized[:5]